CA2505957A1 - Method and system for classifying display pages using summaries - Google Patents
Method and system for classifying display pages using summaries Download PDFInfo
- Publication number
- CA2505957A1 CA2505957A1 CA002505957A CA2505957A CA2505957A1 CA 2505957 A1 CA2505957 A1 CA 2505957A1 CA 002505957 A CA002505957 A CA 002505957A CA 2505957 A CA2505957 A CA 2505957A CA 2505957 A1 CA2505957 A1 CA 2505957A1
- Authority
- CA
- Canada
- Prior art keywords
- summarization
- sentence
- technique
- web page
- summarization technique
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/34—Browsing; Visualisation therefor
- G06F16/345—Summarisation for human users
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/951—Indexing; Web crawling techniques
Abstract
A method and system for classifying display pages based on automatically generated summaries of display pages. A web page classification system uses a web page summarization system to generate summaries of web pages. The summary of a web page may include the sentences of the web page that are most closely related to the primary topic of the web page. The summarization system may combine the benefits of multiple summarization techniques to identify the sentences of a web page that represent the primary topic of the web page. Once the summary is generated, the classification system may apply conventional classification techniques to the summary to classify the web page. The classification system may use conventional classification techniques such as a Naïve Bayesian classifier or a support vector machine to identify the classifications of a web page based on the summary generated by the summarization system.
Description
METHOD AND SYSTEM FOR CLASSIFYING DISPLAY PAGES
USING SUMMARIES
TECHNICAL FIELD
The described technology relates generally to automatically classifying s information.
BACKGROUND
Many search engine services, such as Google and Overture, provide for searching for information that is accessible via the Internet. These search engine services allow users to search for display pages, such as web pages, that may be of interest to users. After a user submits a search request that includes search terms, the search engine service identifies web pages that may be related to those search terms. To quickly identify related web pages, the search engine services may maintain a mapping of keywords to web pages. This mapping may be generated by "crawling" the web (i.e., the World Wide Web) to identify the keywords of each web page. To crawl the web, a search engine service may use a list of root web pages to identify all web pages that are accessible through those root web pages. The keywords of any particular web page can be identified using various well-known information retrieval techniques, such as identifying the words of a headline, the words supplied in the metadata of the web page, the words that
USING SUMMARIES
TECHNICAL FIELD
The described technology relates generally to automatically classifying s information.
BACKGROUND
Many search engine services, such as Google and Overture, provide for searching for information that is accessible via the Internet. These search engine services allow users to search for display pages, such as web pages, that may be of interest to users. After a user submits a search request that includes search terms, the search engine service identifies web pages that may be related to those search terms. To quickly identify related web pages, the search engine services may maintain a mapping of keywords to web pages. This mapping may be generated by "crawling" the web (i.e., the World Wide Web) to identify the keywords of each web page. To crawl the web, a search engine service may use a list of root web pages to identify all web pages that are accessible through those root web pages. The keywords of any particular web page can be identified using various well-known information retrieval techniques, such as identifying the words of a headline, the words supplied in the metadata of the web page, the words that
2 o are highlighted, and so on. The search engine service may generate a relevance score to indicate how relevant the information of the web page may be to the search request based on the closeness of each match, web page popularity (e.g., Google's PageRank), and so on. The search engine service then displays to the user links to those web pages in an order that is based on their rankings.
2 s Although search engine services may return many web pages as a search result, the presenting of the web pages in rank order may make it difficult for a user to actually find those web pages of particular interest to the user.
Since the web pages that are presented first may be directed to popular topics, a user who is interested in an obscure topic may need to scan many pages of the search
2 s Although search engine services may return many web pages as a search result, the presenting of the web pages in rank order may make it difficult for a user to actually find those web pages of particular interest to the user.
Since the web pages that are presented first may be directed to popular topics, a user who is interested in an obscure topic may need to scan many pages of the search
3 o result to find a web page of interest. To make it easier for a user to find web pages of interest, the web pages of a search result could be presented in a hierarchical organization based on some classification or categorization of the web pages. For example, if a user submits a search request of "court battles," the search result may contain web pages that can be classified as sports-related or s legal-related. The user may prefer to be presented initially with a list of classifications of the web pages so that the user can select the classification of web pages that is of interest. For example, the user might be first presented with an indication that the web pages of the search result have been classified as sports-related and legal-related. The user can then select the legal-related io classification to view web pages that are legal-related. In contrast, since sports web pages are more popular than legal web pages, a user might have to scan many pages to find legal-related web pages if the most popular web pages are presented first.
It would be impractical to manually classify the millions of web pages that Z5 are currently available. Although automated classification techniques have been used to classify text-based content, those techniques are not generally applicable to the classification of web pages. Web pages have an organization that includes noisy content, such as an advertisement or a navigation bar, that is not directly related to the primary topic of the web page. Because conventional text-based 2 o classification techniques would use such noisy content when classifying a web page, these techniques would tend to produce incorrect classifications of web pages.
It would be desirable to have a classification technique for web pages that would base the classification of a web page on the primary topic of the web page 25 and give little weight to noisy content of the web page.
SUMMARY
A classification and summarization system classifies display pages such as web pages based on automatically generated summaries of the display pages. In one embodiment, a web page classification system uses a web page 3 o summarization system to generate summaries of web pages. The summary of a web page may include the sentences of the web page that are most closely related to the primary topic of the web page. The summarization system may combine the benefits of multiple summarization techniques to identify the sentences of a web page that represent the primary topic of the web page. Once a summary is generated, the classification system may apply conventional classification techniques to the summary to classify the web page.
BRIEF DESCRIPTION OF THE DRAWINGS
s Figure 1 is block diagram that illustrates components of a classification system and a summarization system in one embodiment.
Figure 2 is a flow diagram that illustrates the processing of the classify web page component in one embodiment.
Figure 3 is a flow diagram that illustrates the processing of the summarize 1 o web page component in one embodiment.
Figure 4 is a flow diagram that illustrates the processing of the calculate scores component in one embodiment.
Figure 5 is a flow diagram that illustrates the processing of the calculate Luhn score component in one embodiment.
15 Figure 6 is a flow diagram that illustrates the processing of the calculate latent semantic analysis score component in one embodiment.
Figure 7 is a flow diagram that illustrates the processing of the calculate content body score component in one embodiment.
Figure 8 is a flow diagram that illustrates the processing of the calculate 2 o supervised score component in one embodiment.
Figure 9 is a flow diagram that illustrates the processing of the combine scores component in one embodiment.
DETAILED DESCRIPTION
A method and system for classifying display pages based on automatically 2 s generated summaries of display pages is provided. In one embodiment, a web page classification system uses a web page summarization system to generate summaries of web pages. The summary of a web page may include the sentences of the web page that are most closely related to the primary topic of the web page. Once the summary is generated, the classification system may apply 3 o conventional classification techniques to the summary to classify the web page.
The summarization system may combine the benefits of multiple summarization techniques to identify the sentences of a web page that represent the primary topic of the web page. In one embodiment, the summarization system uses a Luhn summarization technique, a latent semantic analysis summarization technique, a content body summarization technique, and a supervised summarization technique either individually or in combination to generate a s summary. The summarization system uses each of the summarization techniques to generate a summarization technique-specific score for each sentence of a web page. The summarization system then combines the summarization technique-specific scores for a sentence to generate an overall score for that sentence.
The summarization system selects the sentences of the web page with the highest overall scores to form the summary of the web page. The classification system may use conventional classification techniques such as a Naive Bayesian classifier or a support vector machine to identify the classifications of a web page based on the summary generated by the summarization system. In this way, web pages can be automatically classified based on automatically generated i5 summaries of the web pages.
In one embodiment, the summarization system uses a modified version of the Luhn summarization technique to generate a Luhn score for each sentence of a web page. The Luhn summarization technique generates a score for a sentence that is based on the "significant words" that are in the sentence. To 2 o generate a score for a sentence, the Luhn summarization technique identifies a portion of the sentence that is bracketed by significant words that are not more than a certain number of non-significant words apart. The Luhn summarization technique calculates the score of the sentence as the ratio of the square of the number of significant words contained in the bracketed portion divided by the 25 number of words within the bracketed portion. (See H.P. Luhn, The Automatic Creation of Literature Abstracts, 2 IBM J. of RES. & DEV. No. 2, 159-65 (April 1958).) The summarization system modifies the Luhn summarization technique by defining a collection of significant words for each classification. For example, a sports-related classification may have a collection of significant words that 3 o includes "court," "basketball," and "sport," whereas a legal-related classification may have a collection of significant words that includes "court," "attorney,"
and "criminal." The summarization system may identify the collections of significant words based on a training set of web pages that have been pre-classified. The summarization system may select the most frequently used words on the web
It would be impractical to manually classify the millions of web pages that Z5 are currently available. Although automated classification techniques have been used to classify text-based content, those techniques are not generally applicable to the classification of web pages. Web pages have an organization that includes noisy content, such as an advertisement or a navigation bar, that is not directly related to the primary topic of the web page. Because conventional text-based 2 o classification techniques would use such noisy content when classifying a web page, these techniques would tend to produce incorrect classifications of web pages.
It would be desirable to have a classification technique for web pages that would base the classification of a web page on the primary topic of the web page 25 and give little weight to noisy content of the web page.
SUMMARY
A classification and summarization system classifies display pages such as web pages based on automatically generated summaries of the display pages. In one embodiment, a web page classification system uses a web page 3 o summarization system to generate summaries of web pages. The summary of a web page may include the sentences of the web page that are most closely related to the primary topic of the web page. The summarization system may combine the benefits of multiple summarization techniques to identify the sentences of a web page that represent the primary topic of the web page. Once a summary is generated, the classification system may apply conventional classification techniques to the summary to classify the web page.
BRIEF DESCRIPTION OF THE DRAWINGS
s Figure 1 is block diagram that illustrates components of a classification system and a summarization system in one embodiment.
Figure 2 is a flow diagram that illustrates the processing of the classify web page component in one embodiment.
Figure 3 is a flow diagram that illustrates the processing of the summarize 1 o web page component in one embodiment.
Figure 4 is a flow diagram that illustrates the processing of the calculate scores component in one embodiment.
Figure 5 is a flow diagram that illustrates the processing of the calculate Luhn score component in one embodiment.
15 Figure 6 is a flow diagram that illustrates the processing of the calculate latent semantic analysis score component in one embodiment.
Figure 7 is a flow diagram that illustrates the processing of the calculate content body score component in one embodiment.
Figure 8 is a flow diagram that illustrates the processing of the calculate 2 o supervised score component in one embodiment.
Figure 9 is a flow diagram that illustrates the processing of the combine scores component in one embodiment.
DETAILED DESCRIPTION
A method and system for classifying display pages based on automatically 2 s generated summaries of display pages is provided. In one embodiment, a web page classification system uses a web page summarization system to generate summaries of web pages. The summary of a web page may include the sentences of the web page that are most closely related to the primary topic of the web page. Once the summary is generated, the classification system may apply 3 o conventional classification techniques to the summary to classify the web page.
The summarization system may combine the benefits of multiple summarization techniques to identify the sentences of a web page that represent the primary topic of the web page. In one embodiment, the summarization system uses a Luhn summarization technique, a latent semantic analysis summarization technique, a content body summarization technique, and a supervised summarization technique either individually or in combination to generate a s summary. The summarization system uses each of the summarization techniques to generate a summarization technique-specific score for each sentence of a web page. The summarization system then combines the summarization technique-specific scores for a sentence to generate an overall score for that sentence.
The summarization system selects the sentences of the web page with the highest overall scores to form the summary of the web page. The classification system may use conventional classification techniques such as a Naive Bayesian classifier or a support vector machine to identify the classifications of a web page based on the summary generated by the summarization system. In this way, web pages can be automatically classified based on automatically generated i5 summaries of the web pages.
In one embodiment, the summarization system uses a modified version of the Luhn summarization technique to generate a Luhn score for each sentence of a web page. The Luhn summarization technique generates a score for a sentence that is based on the "significant words" that are in the sentence. To 2 o generate a score for a sentence, the Luhn summarization technique identifies a portion of the sentence that is bracketed by significant words that are not more than a certain number of non-significant words apart. The Luhn summarization technique calculates the score of the sentence as the ratio of the square of the number of significant words contained in the bracketed portion divided by the 25 number of words within the bracketed portion. (See H.P. Luhn, The Automatic Creation of Literature Abstracts, 2 IBM J. of RES. & DEV. No. 2, 159-65 (April 1958).) The summarization system modifies the Luhn summarization technique by defining a collection of significant words for each classification. For example, a sports-related classification may have a collection of significant words that 3 o includes "court," "basketball," and "sport," whereas a legal-related classification may have a collection of significant words that includes "court," "attorney,"
and "criminal." The summarization system may identify the collections of significant words based on a training set of web pages that have been pre-classified. The summarization system may select the most frequently used words on the web
4 pages with a certain classification as the collection of significant words for that classification. The summarization system may also remove certain stop words from the collection that may represent noisy content. When scoring a sentence of a web page, the modified Luhn summarization technique calculates a score for each classification. The summarization technique then averages the scores for each classification that are above a threshold level to give a combined Luhn score for the sentence. The summarization system may select the sentences with the highest Luhn scores to form the summary.
In one embodiment, the summarization system uses a latent semantic to analysis summarization technique to generate a latent semantic analysis score for each sentence of a web page. The latent semantic analysis summarization technique uses singular value decomposition to generate a score for each sentence. The summarization system generates a word-sentence matrix for the web page that contains a weighted term-frequency value for each word-sentence combination. The matrix may be represented by the following:
A=UEVT (1 ) where A represents the word-sentence matrix, U is a column-orthonormal matrix whose columns are left singular vectors, E is a diagonal matrix whose diagonal elements are non-negative singular values sorted in descending order, and V is an orthonormal matrix whose columns are right singular vectors. After 2 o decomposing the matrix into U , E , and V , the summarization system uses the right singular vectors to generate the scores for the sentences. (See Y.H.
Gong &
X. Liu, Generic Text Summarization Using Relevance Measure and Latent Semantic AnaIySIS, In PROC. OF THE 24T" ANNUAL INTERNATIONAL ACM SIGIR, NeW
Orleans, Louisiana, 19-25 (2001 ).) The summarization system may select the first right singular vector and select the sentence that has the highest index value within that vector. The summarization system then gives that sentence the highest score. The summarization system then selects the second right singular vector and gives the sentence that has the highest index value within that vector the second highest score. The summarization system then continues in a similar 3 o manner to generate the scores for the other sentences. The summarization system may select the sentences with the highest scores to form the summary of the web page.
In one embodiment, the summarization system uses a latent semantic to analysis summarization technique to generate a latent semantic analysis score for each sentence of a web page. The latent semantic analysis summarization technique uses singular value decomposition to generate a score for each sentence. The summarization system generates a word-sentence matrix for the web page that contains a weighted term-frequency value for each word-sentence combination. The matrix may be represented by the following:
A=UEVT (1 ) where A represents the word-sentence matrix, U is a column-orthonormal matrix whose columns are left singular vectors, E is a diagonal matrix whose diagonal elements are non-negative singular values sorted in descending order, and V is an orthonormal matrix whose columns are right singular vectors. After 2 o decomposing the matrix into U , E , and V , the summarization system uses the right singular vectors to generate the scores for the sentences. (See Y.H.
Gong &
X. Liu, Generic Text Summarization Using Relevance Measure and Latent Semantic AnaIySIS, In PROC. OF THE 24T" ANNUAL INTERNATIONAL ACM SIGIR, NeW
Orleans, Louisiana, 19-25 (2001 ).) The summarization system may select the first right singular vector and select the sentence that has the highest index value within that vector. The summarization system then gives that sentence the highest score. The summarization system then selects the second right singular vector and gives the sentence that has the highest index value within that vector the second highest score. The summarization system then continues in a similar 3 o manner to generate the scores for the other sentences. The summarization system may select the sentences with the highest scores to form the summary of the web page.
5 In one embodiment, the summarization system uses a content body summarization technique to generate a content body score for each sentence of a web page. The content body summarization technique identifies the content body of a web page and gives a high score to the sentences within the content body.
s To identify the content body of a web page, the content body summarization technique identifies basic objects and composite objects of the web page. A
basic object is the smallest information area that cannot be further divided. For example, in HTML, a basic object is a non-breakable element within two tags or an embedded object. A composite object is a set of basic objects or other 1 o composite objects that combine to perform a function. After identifying the objects, the summarization system categorizes the objects into categories such as information, navigation, interaction, decoration, or special function. The information category is for objects that present content information, the navigation category is for objects that present a navigation guide, the interaction category is 15 for objects that present user interactions (e.g., input field), the decoration category is for objects that present decorations, and a special function category is for objects that present information such as legal information, contact information, logo information, and so on. (See J.L. Chen, et al., Function-based Object Model Towards Website Adaptation, PFtoc. of WWW10, Hong Kong, China (2001).) In 2 0 one embodiment, the summarization system builds a term frequency by inverted document frequency index (i.e., TF*IDF) for each object. The summarization system then calculates the similarity between pairs of objects using a similarity computation such as cosine similarity. If the similarity between the objects of the pair is greater than a threshold level, the summarization system links the objects 2 s of the pair. The summarization system then identifies the object that has the most links to it as the core object that represents the primary topic of the web page.
The content body of the web page is the core object along with each object that has a link to the core object. The summarization system gives a high score to each sentence of the content body and a low score to every other sentence of the 3 o web page. The summarization system may select the sentences with a high score to form the summary of the web page.
In one embodiment, the summarization system uses a supervised summarization technique to generate a supervised score for each sentence of a web page. The supervised summarization technique uses training data to learn a
s To identify the content body of a web page, the content body summarization technique identifies basic objects and composite objects of the web page. A
basic object is the smallest information area that cannot be further divided. For example, in HTML, a basic object is a non-breakable element within two tags or an embedded object. A composite object is a set of basic objects or other 1 o composite objects that combine to perform a function. After identifying the objects, the summarization system categorizes the objects into categories such as information, navigation, interaction, decoration, or special function. The information category is for objects that present content information, the navigation category is for objects that present a navigation guide, the interaction category is 15 for objects that present user interactions (e.g., input field), the decoration category is for objects that present decorations, and a special function category is for objects that present information such as legal information, contact information, logo information, and so on. (See J.L. Chen, et al., Function-based Object Model Towards Website Adaptation, PFtoc. of WWW10, Hong Kong, China (2001).) In 2 0 one embodiment, the summarization system builds a term frequency by inverted document frequency index (i.e., TF*IDF) for each object. The summarization system then calculates the similarity between pairs of objects using a similarity computation such as cosine similarity. If the similarity between the objects of the pair is greater than a threshold level, the summarization system links the objects 2 s of the pair. The summarization system then identifies the object that has the most links to it as the core object that represents the primary topic of the web page.
The content body of the web page is the core object along with each object that has a link to the core object. The summarization system gives a high score to each sentence of the content body and a low score to every other sentence of the 3 o web page. The summarization system may select the sentences with a high score to form the summary of the web page.
In one embodiment, the summarization system uses a supervised summarization technique to generate a supervised score for each sentence of a web page. The supervised summarization technique uses training data to learn a
6 summarize function that identifies whether a sentence should be selected as part of a summary. The supervised summarization technique represents each sentence by a feature vector. In one embodiment, the supervised summarization technique uses the features defined in Table 1 where f;~ represents the value of the ith feature of sentence i.
Table 1 Feature Description___ fil the position of a sentence Si in its containing paragraph.
fit the length of a sentence Si which is the number of words in Si.
fi3 E TFW*SFW, which takes into account not only the number of words w into consideration, but also its distribution among sentences where TFW
is the number of occurrences of word w in a target web page and where SFW
is the number of sentences including the word w in the target web page.
fi4 the similarity between Si and the title, which may be calculated as the dot product between the sentence and the title.
fis the cosine similarity between Si and all text in the web page.
fib the cosine similarity between Si and metadata of the web page.
fi' the number of occurrences of a word from a special word set that are in Si.
The special word set may be built by collecting the words in the web page that are highlighted (e. g., italicized, bold faced, or under-lined).
fib the average font size of the words in Si. In general, larger font size in a web page is given higher importance.
The summarization system may use a Naive Bayesian classifier to learn the summarize function. The summarize function can be represented by the following:
_ ~8= P(.f; ~ S E S)P(S E S) (2) P (S E S ~ .ft ~ fz ~.. f8 ) - ~
1 1;_, p(.f; ) where p(s E S) stands for the compression rate of the summarizer (which can be predefined for different applications), p( f~ ) is the probability of each feature j , and p(f~ ~ s E S) is the conditional probability of each feature j . The latter two factors can be estimated from the training set.
In one embodiment, the summarization system combines the scores of the to Luhn summarization technique, the latent semantic analysis summarization s technique, the content body summarization technique, and the supervised summarization technique to generate an overall score. The scores may be combined as follows:
'S - '~luhn + '~Isa +'Scb + 'sup where S represents the combined score, Slubn represents the Luhn score, Slso represents the latent semantic analysis score, S~b represents the content body score, and Scup represents the supervised score. Alternatively, the summarization system may apply a weighting factor to each summarization technique score so that not all the summarization technique scores are weighted equally. For example, if the Luhn score is thought to be a more accurate reflection of the 1 o relatedness of a sentence to the primary topic of the web page, then the weighting factor for the Luhn score might be .7 and the weighting factor for the other scores might be .1 each. If a weighting factor for a summarization technique is set to zero, then the summarization system does not use that summarization technique.
One skilled in the art will appreciate that any number of the summarization 1 s techniques can have their weights set to zero. For example, if a weighting factor of 1 is used for the Luhn score and for zero for the other scores, then the "combined" score would be simply the Luhn score. In addition, the summarization system may normalize each of the summarization technique scores. The summarization system may also use a non-linear combination of the 2 o summarization technique scores. The summarization system may select the sentences with the highest combined scores to form the summary of the web page.
In one embodiment, the classification system uses a Naive Bayesian classifier to classify a web page based on its summary. The Naive Bayesian 2 5 classifier uses Bayes' rule, which may be defined as follows:
N(wk,d;
P(C~ Ie)1~k 1 P(wk (C~~e ~
P \C% ~di o~ - N(wk,di~ 't ~~r--1 P\Cr B/llk=1 P\Wk Cr~e where P(c~ I dl;B) can be calculated by counting the frequency with each category c~ occurring in the training data, C is the number of categories, p(wl ~ c~) is a probability that word w; occurs in class c~ , N(wk,d;) is the number of occurrences of a word wk in d; , and n is the number of words in the training data. (See A.
McCallum & K. Nigam, A Comparison of Event Models for Naive Bayes Text Classification, In AAAI-98 WORKSHOP ON LEARNING FOR TEXT CATEGORIZATION
(1998).) Since w; may be small in the training data, a Laplace smoothing may be s used to estimate its value.
In an alternate embodiment the classification system uses a support vector machine to classify a web page based on its summary. A support vector machine operates by finding a hyper-surface in the space of possible inputs. The hyper-surface attempts to split the positive examples from the negative examples by 1 o maximizing the distance between the nearest of the positive and negative examples to the hyper-surface. This allows for correct classification of data that is similar to but not identical to the training data. Various techniques can be used to train a support vector machine. One technique uses a sequential minimal optimization algorithm that breaks the large quadratic programming problem down 15 into a series of small quadratic programming problems that can be solved analytically. (See Sequential Minimal Optimization, at http://research.micro-soft.com/~jplatt/smo.html. ) Figure 1 is block diagram that illustrates components of a classification system and a summarization system in one embodiment. The classification 2 o system 110 includes a classify web page component 111 and a classifier component 112. The summarization system 120 includes a summarize web page component 121, a sort sentences component 122, a calculate scores component 123, and a select top sentences component 124. The classify web page component uses the summarize web page component to generate a summary for 2 s a web page and then uses the classifier component to classify the web page based on the summary. The summarize web page component uses the calculate scores component to calculate a score for each sentence of the web page. The summarize web page component then uses the sort sentences component to sort the sentences of the web page based on their scores and the select top sentences 3 o component to select the sentences with the highest scores to form the summary of the web page. The calculate scores component uses a calculate Luhn score component 125, a calculate latent semantic analysis score component 126, a calculate content body score component 127, and a calculate supervised score component 128 to generate scores for various summarization techniques. The to calculate scores component then combines the scores for the summarization techniques to provide an overall score for each sentence.
The computing device on which the summarization system is implemented may include a central processing unit, memory, input devices (e.g., keyboard and s pointing devices), output devices (e.g., display devices), and storage devices (e.g., disk drives). The memory and storage devices are computer-readable media that may contain instructions that implement the summarization system.
In addition, the data structures and message structures may be stored or transmitted via a data transmission medium, such as a signal on a communications link.
to Various communications links may be used, such as the Internet, a local area network, a wide area network, or a point-to-point dial-up connection.
The summarization system may be implemented in various operating environments. The operating environment described herein is only one example of a suitable operating environment and is not intended to suggest any limitation 15 as to the scope of use or functionality of the summarization system. Other well-known computing systems, environments, and configurations that may be suitable for use include personal computers, server computers, hand-held or laptop devices, multiprocessor systems, microprocessor-based systems, programmable consumer electronics, network PCs, minicomputers, mainframe computers, 2 o distributed computing environments that include any of the above systems or devices, and the like.
The summarization system may be described in the general context of computer-executable instructions, such as program modules, executed by one or more computers or other devices. Generally, program modules include routines, 25 programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. Typically, the functionality of the program modules may be combined or distributed as desired in various embodiments.
Figure 2 is a flow diagram that illustrates the processing of the classify web 3 o page component in one embodiment. The component is passed a web page and returns its classifications. In block 201, the component invokes the summarize web page component to generate a summary for the web page. In block 202, the component classifies the web page based on the summary of the web page using m a classifier such as a Naive Bayesian classifier or a support vector machine.
The component then completes.
Figure 3 is a flow diagram that illustrates the processing of the summarize web page component in one embodiment. The component is passed a web page, calculates a score for each sentence of the web page, and selects the sentences with the highest scores to form the summary of the web page. In block 301, the component invokes the calculate scores component to calculate a score for each sentence. In block 302, the component sorts the sentences based on the calculated scores. In block 303, the component selects the sentences with the top 1 o scores to form the summary for the web page. The component then returns the summary.
Figure 4 is a flow diagram that illustrates the processing of the calculate scores component in one embodiment. The component is passed a web page, calculates various summarization technique scores for the sentences of the web page, and calculates a combined score for each sentence based on those summarization technique scores. The component may alternatively calculate a score using only one summarization technique or various combinations of the summarization techniques. In block 401, the component invokes the calculate Luhn score component to calculate a Luhn score for each sentence of the web 2 o page. In block 402, the component invokes the calculate latent semantic analysis score component to calculate a latent semantic analysis score for each sentence of the web page. In block 403, the component invokes the calculate content body score component to calculate a content body score for each sentence of the web page. In block 404, the component invokes the calculate supervised score component to calculate a supervised score for each sentence of the web page.
In block 405, the component invokes a combine scores component to calculate a combined score for each sentence of the web page. The component then returns the combined scores.
Figure 5 is a flow diagram that illustrates the processing of the calculate 3 o Luhn score component in one embodiment. The component is passed a web page and calculates a Luhn score for each sentence of the passed web page. In block 501, the component selects the next sentence of the web page. In decision block 502, if all the sentences of the web page have already been selected, then the component returns the Luhn scores, else the component continues at block 503. In blocks 503-509, the component loops generating a class score for the selected sentence for each classification. In block 503, the component selects the next classification. In decision block 504, if all the classifications have already been selected, then the component continues at block 510, else the component s continues at block 505. In block 505, the component identifies words of the selected sentence that are bracketed by significant words of the selected classification. In decision block 506, if bracketed words are identified, then the component continues at block 507, else the component loops to block 503 to select the next classification. In block 507, the component counts the significant to words within the bracketed portion of the selected sentence. In block 508, the component counts the words within the bracketed portion of the selected sentence. In block 509, the component calculates a score for the classification as the square of the count of significant words divided by the count of words.
The component then loops to block 503 to select the next classification. In block 510, is the component calculates the Luhn score for the selected sentence as a sum of the class scores divided by the number of classifications for which a bracketed portion of the selected sentence was identified (i.e., the average of the class scores that were calculated). The component then loops to block 501 to select the next sentence.
2 o Figure 6 is a flow diagram that illustrates the processing of the calculate latent semantic analysis score component in one embodiment. The component is passed a web page and calculates a latent semantic analysis score for each sentence of the passed web page. In blocks 601-603, the component loops constructing a term-by-weight vector for each sentence of the web page. In block 2 s 601, the component selects the next sentence of the web page. In decision block 602, if all the sentences of the web page have already been selected, then the component continues at block 604, else the component continues at block 603.
In block 603, the component constructs a term-by-weight vector for the selected sentence and then loops to block 601 to select the next sentence. The term-by-3 o weight vectors for the sentences form a matrix that is decomposed to give a matrix of right singular vectors. In block 604, the component performs singular value decomposition of that matrix to generate the right singular vectors. In blocks 607, the component loops setting a score for each sentence based on the right singular vectors. In block 605, the component selects the next right singular vector. In decision block 606, if all the right singular vectors have already been selected, then the component returns the scores as the latent semantic analysis scores, else the component continues at block 607. In block 607, the component sets the score of the sentence with the highest index value of the selected right s singular vector and then loops to block 605 to select the next right singular vector.
Figure 7 is a flow diagram that illustrates the processing of the calculate content body score component in one embodiment. The component is passed a web page and calculates a content body score for each sentence of the passed web page. In block 701, the component identifies the basic objects of the web to page. In block 702, the component identifies the composite objects of the web page. In blocks 703-705, the component loops generating a term frequency/inverted document frequency vector for each object. In block 703, the component selects the next object. In decision block 704, if all the objects have already been selected, then the component continues at block 706, else the 15 component continues at block 705. In block 705, the component generates the term frequency/inverted document frequency vector for the selected object and then loops to block 703 to select the next object. In blocks 706-710, the component loops calculating the similarity between pairs of objects. In block 706, the component selects the next pair of objects. In decision block 707, if all the 2 o pairs of objects have already been selected, then the component continues at block 711, else the component continues at block 708. In block 708, the component calculates the similarity between the selected pair of objects. In decision block 709, if the similarity is higher than a threshold level of similarity, then the component continues at block 710, else the component loops to block 2s 706 to select the next pair of objects. In block 710, the component adds a link between the selected pair of objects and then loops to block 706 to select the next pair of objects. In blocks 711-715, the component identifies the content body of the web page by identifying a core object and all objects with links to that core object. In block 711, the component identifies the core object as the object with 3o the greatest number of links to it. In block 712, the component selects the next sentence of the web page. In decision block 713, if all the sentences have already been selected, then the component returns the content body scores, else the component continues at block 714. In decision block 714, if the sentence is within an object that is linked to the core object, then the sentence is within the content body and the component continues at block 715, else the component sets the score of the selected sentence to zero and loops to block 712 to select the next sentence. In block 715, the component sets the score of the selected sentence to a high score and then loops to block 712 to select the next sentence.
Figure 8 is a flow diagram that illustrates the processing of the calculate supervised score component in one embodiment. The component is passed a web page and calculates a supervised score for each sentence of the web page.
In block 801, the component selects the next sentence of the web page. In decision block 802, if all the sentences of the web page have already been 1 o selected, then the component returns the supervised scores, else the component continues at block 803. In block 803, the component generates the feature vector for the selected sentence. In block 804, the component calculates the score for the selected sentence using the generated feature vector and the learned summarize function. The component then loops to block 801 to select the next sentence.
Figure 9 is a flow diagram that illustrates the processing of the combine scores component in one embodiment. The component generates a combined score for each sentence of a web page based on the Luhn score, the latent semantic analysis score, the content body score, and the supervised score. In 2 o block 901, the component selects the next sentence of the web page. In decision block 902, if all the sentences have already been selected, then the component returns the combined scores, else the component continues at block 903. In block 903, the component combines the scores for the selected sentence and then loops to block 901 to select the next sentence.
One skilled in the art will appreciate that although specific embodiments of the summarization system have been described herein for purposes of illustration, various modifications may be made without deviating from the spirit and scope of the invention. One skilled in the art will appreciate that classification refers to the process of identifying the class or category associated with a display page.
The 3 o classes may be predefined. The attributes of a display page to be classified may be compared to attributes derived from other display pages that have been classified (e.g., a training set). Based on the comparison, the display page is classified into the class whose display page attributes are similar to those of the display page being classified. Clustering, in contrast, refers to the process of identifying from a set of display pages groups of display pages that are similar to each other. Accordingly, the invention is not limited except by the appended claims.
Table 1 Feature Description___ fil the position of a sentence Si in its containing paragraph.
fit the length of a sentence Si which is the number of words in Si.
fi3 E TFW*SFW, which takes into account not only the number of words w into consideration, but also its distribution among sentences where TFW
is the number of occurrences of word w in a target web page and where SFW
is the number of sentences including the word w in the target web page.
fi4 the similarity between Si and the title, which may be calculated as the dot product between the sentence and the title.
fis the cosine similarity between Si and all text in the web page.
fib the cosine similarity between Si and metadata of the web page.
fi' the number of occurrences of a word from a special word set that are in Si.
The special word set may be built by collecting the words in the web page that are highlighted (e. g., italicized, bold faced, or under-lined).
fib the average font size of the words in Si. In general, larger font size in a web page is given higher importance.
The summarization system may use a Naive Bayesian classifier to learn the summarize function. The summarize function can be represented by the following:
_ ~8= P(.f; ~ S E S)P(S E S) (2) P (S E S ~ .ft ~ fz ~.. f8 ) - ~
1 1;_, p(.f; ) where p(s E S) stands for the compression rate of the summarizer (which can be predefined for different applications), p( f~ ) is the probability of each feature j , and p(f~ ~ s E S) is the conditional probability of each feature j . The latter two factors can be estimated from the training set.
In one embodiment, the summarization system combines the scores of the to Luhn summarization technique, the latent semantic analysis summarization s technique, the content body summarization technique, and the supervised summarization technique to generate an overall score. The scores may be combined as follows:
'S - '~luhn + '~Isa +'Scb + 'sup where S represents the combined score, Slubn represents the Luhn score, Slso represents the latent semantic analysis score, S~b represents the content body score, and Scup represents the supervised score. Alternatively, the summarization system may apply a weighting factor to each summarization technique score so that not all the summarization technique scores are weighted equally. For example, if the Luhn score is thought to be a more accurate reflection of the 1 o relatedness of a sentence to the primary topic of the web page, then the weighting factor for the Luhn score might be .7 and the weighting factor for the other scores might be .1 each. If a weighting factor for a summarization technique is set to zero, then the summarization system does not use that summarization technique.
One skilled in the art will appreciate that any number of the summarization 1 s techniques can have their weights set to zero. For example, if a weighting factor of 1 is used for the Luhn score and for zero for the other scores, then the "combined" score would be simply the Luhn score. In addition, the summarization system may normalize each of the summarization technique scores. The summarization system may also use a non-linear combination of the 2 o summarization technique scores. The summarization system may select the sentences with the highest combined scores to form the summary of the web page.
In one embodiment, the classification system uses a Naive Bayesian classifier to classify a web page based on its summary. The Naive Bayesian 2 5 classifier uses Bayes' rule, which may be defined as follows:
N(wk,d;
P(C~ Ie)1~k 1 P(wk (C~~e ~
P \C% ~di o~ - N(wk,di~ 't ~~r--1 P\Cr B/llk=1 P\Wk Cr~e where P(c~ I dl;B) can be calculated by counting the frequency with each category c~ occurring in the training data, C is the number of categories, p(wl ~ c~) is a probability that word w; occurs in class c~ , N(wk,d;) is the number of occurrences of a word wk in d; , and n is the number of words in the training data. (See A.
McCallum & K. Nigam, A Comparison of Event Models for Naive Bayes Text Classification, In AAAI-98 WORKSHOP ON LEARNING FOR TEXT CATEGORIZATION
(1998).) Since w; may be small in the training data, a Laplace smoothing may be s used to estimate its value.
In an alternate embodiment the classification system uses a support vector machine to classify a web page based on its summary. A support vector machine operates by finding a hyper-surface in the space of possible inputs. The hyper-surface attempts to split the positive examples from the negative examples by 1 o maximizing the distance between the nearest of the positive and negative examples to the hyper-surface. This allows for correct classification of data that is similar to but not identical to the training data. Various techniques can be used to train a support vector machine. One technique uses a sequential minimal optimization algorithm that breaks the large quadratic programming problem down 15 into a series of small quadratic programming problems that can be solved analytically. (See Sequential Minimal Optimization, at http://research.micro-soft.com/~jplatt/smo.html. ) Figure 1 is block diagram that illustrates components of a classification system and a summarization system in one embodiment. The classification 2 o system 110 includes a classify web page component 111 and a classifier component 112. The summarization system 120 includes a summarize web page component 121, a sort sentences component 122, a calculate scores component 123, and a select top sentences component 124. The classify web page component uses the summarize web page component to generate a summary for 2 s a web page and then uses the classifier component to classify the web page based on the summary. The summarize web page component uses the calculate scores component to calculate a score for each sentence of the web page. The summarize web page component then uses the sort sentences component to sort the sentences of the web page based on their scores and the select top sentences 3 o component to select the sentences with the highest scores to form the summary of the web page. The calculate scores component uses a calculate Luhn score component 125, a calculate latent semantic analysis score component 126, a calculate content body score component 127, and a calculate supervised score component 128 to generate scores for various summarization techniques. The to calculate scores component then combines the scores for the summarization techniques to provide an overall score for each sentence.
The computing device on which the summarization system is implemented may include a central processing unit, memory, input devices (e.g., keyboard and s pointing devices), output devices (e.g., display devices), and storage devices (e.g., disk drives). The memory and storage devices are computer-readable media that may contain instructions that implement the summarization system.
In addition, the data structures and message structures may be stored or transmitted via a data transmission medium, such as a signal on a communications link.
to Various communications links may be used, such as the Internet, a local area network, a wide area network, or a point-to-point dial-up connection.
The summarization system may be implemented in various operating environments. The operating environment described herein is only one example of a suitable operating environment and is not intended to suggest any limitation 15 as to the scope of use or functionality of the summarization system. Other well-known computing systems, environments, and configurations that may be suitable for use include personal computers, server computers, hand-held or laptop devices, multiprocessor systems, microprocessor-based systems, programmable consumer electronics, network PCs, minicomputers, mainframe computers, 2 o distributed computing environments that include any of the above systems or devices, and the like.
The summarization system may be described in the general context of computer-executable instructions, such as program modules, executed by one or more computers or other devices. Generally, program modules include routines, 25 programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. Typically, the functionality of the program modules may be combined or distributed as desired in various embodiments.
Figure 2 is a flow diagram that illustrates the processing of the classify web 3 o page component in one embodiment. The component is passed a web page and returns its classifications. In block 201, the component invokes the summarize web page component to generate a summary for the web page. In block 202, the component classifies the web page based on the summary of the web page using m a classifier such as a Naive Bayesian classifier or a support vector machine.
The component then completes.
Figure 3 is a flow diagram that illustrates the processing of the summarize web page component in one embodiment. The component is passed a web page, calculates a score for each sentence of the web page, and selects the sentences with the highest scores to form the summary of the web page. In block 301, the component invokes the calculate scores component to calculate a score for each sentence. In block 302, the component sorts the sentences based on the calculated scores. In block 303, the component selects the sentences with the top 1 o scores to form the summary for the web page. The component then returns the summary.
Figure 4 is a flow diagram that illustrates the processing of the calculate scores component in one embodiment. The component is passed a web page, calculates various summarization technique scores for the sentences of the web page, and calculates a combined score for each sentence based on those summarization technique scores. The component may alternatively calculate a score using only one summarization technique or various combinations of the summarization techniques. In block 401, the component invokes the calculate Luhn score component to calculate a Luhn score for each sentence of the web 2 o page. In block 402, the component invokes the calculate latent semantic analysis score component to calculate a latent semantic analysis score for each sentence of the web page. In block 403, the component invokes the calculate content body score component to calculate a content body score for each sentence of the web page. In block 404, the component invokes the calculate supervised score component to calculate a supervised score for each sentence of the web page.
In block 405, the component invokes a combine scores component to calculate a combined score for each sentence of the web page. The component then returns the combined scores.
Figure 5 is a flow diagram that illustrates the processing of the calculate 3 o Luhn score component in one embodiment. The component is passed a web page and calculates a Luhn score for each sentence of the passed web page. In block 501, the component selects the next sentence of the web page. In decision block 502, if all the sentences of the web page have already been selected, then the component returns the Luhn scores, else the component continues at block 503. In blocks 503-509, the component loops generating a class score for the selected sentence for each classification. In block 503, the component selects the next classification. In decision block 504, if all the classifications have already been selected, then the component continues at block 510, else the component s continues at block 505. In block 505, the component identifies words of the selected sentence that are bracketed by significant words of the selected classification. In decision block 506, if bracketed words are identified, then the component continues at block 507, else the component loops to block 503 to select the next classification. In block 507, the component counts the significant to words within the bracketed portion of the selected sentence. In block 508, the component counts the words within the bracketed portion of the selected sentence. In block 509, the component calculates a score for the classification as the square of the count of significant words divided by the count of words.
The component then loops to block 503 to select the next classification. In block 510, is the component calculates the Luhn score for the selected sentence as a sum of the class scores divided by the number of classifications for which a bracketed portion of the selected sentence was identified (i.e., the average of the class scores that were calculated). The component then loops to block 501 to select the next sentence.
2 o Figure 6 is a flow diagram that illustrates the processing of the calculate latent semantic analysis score component in one embodiment. The component is passed a web page and calculates a latent semantic analysis score for each sentence of the passed web page. In blocks 601-603, the component loops constructing a term-by-weight vector for each sentence of the web page. In block 2 s 601, the component selects the next sentence of the web page. In decision block 602, if all the sentences of the web page have already been selected, then the component continues at block 604, else the component continues at block 603.
In block 603, the component constructs a term-by-weight vector for the selected sentence and then loops to block 601 to select the next sentence. The term-by-3 o weight vectors for the sentences form a matrix that is decomposed to give a matrix of right singular vectors. In block 604, the component performs singular value decomposition of that matrix to generate the right singular vectors. In blocks 607, the component loops setting a score for each sentence based on the right singular vectors. In block 605, the component selects the next right singular vector. In decision block 606, if all the right singular vectors have already been selected, then the component returns the scores as the latent semantic analysis scores, else the component continues at block 607. In block 607, the component sets the score of the sentence with the highest index value of the selected right s singular vector and then loops to block 605 to select the next right singular vector.
Figure 7 is a flow diagram that illustrates the processing of the calculate content body score component in one embodiment. The component is passed a web page and calculates a content body score for each sentence of the passed web page. In block 701, the component identifies the basic objects of the web to page. In block 702, the component identifies the composite objects of the web page. In blocks 703-705, the component loops generating a term frequency/inverted document frequency vector for each object. In block 703, the component selects the next object. In decision block 704, if all the objects have already been selected, then the component continues at block 706, else the 15 component continues at block 705. In block 705, the component generates the term frequency/inverted document frequency vector for the selected object and then loops to block 703 to select the next object. In blocks 706-710, the component loops calculating the similarity between pairs of objects. In block 706, the component selects the next pair of objects. In decision block 707, if all the 2 o pairs of objects have already been selected, then the component continues at block 711, else the component continues at block 708. In block 708, the component calculates the similarity between the selected pair of objects. In decision block 709, if the similarity is higher than a threshold level of similarity, then the component continues at block 710, else the component loops to block 2s 706 to select the next pair of objects. In block 710, the component adds a link between the selected pair of objects and then loops to block 706 to select the next pair of objects. In blocks 711-715, the component identifies the content body of the web page by identifying a core object and all objects with links to that core object. In block 711, the component identifies the core object as the object with 3o the greatest number of links to it. In block 712, the component selects the next sentence of the web page. In decision block 713, if all the sentences have already been selected, then the component returns the content body scores, else the component continues at block 714. In decision block 714, if the sentence is within an object that is linked to the core object, then the sentence is within the content body and the component continues at block 715, else the component sets the score of the selected sentence to zero and loops to block 712 to select the next sentence. In block 715, the component sets the score of the selected sentence to a high score and then loops to block 712 to select the next sentence.
Figure 8 is a flow diagram that illustrates the processing of the calculate supervised score component in one embodiment. The component is passed a web page and calculates a supervised score for each sentence of the web page.
In block 801, the component selects the next sentence of the web page. In decision block 802, if all the sentences of the web page have already been 1 o selected, then the component returns the supervised scores, else the component continues at block 803. In block 803, the component generates the feature vector for the selected sentence. In block 804, the component calculates the score for the selected sentence using the generated feature vector and the learned summarize function. The component then loops to block 801 to select the next sentence.
Figure 9 is a flow diagram that illustrates the processing of the combine scores component in one embodiment. The component generates a combined score for each sentence of a web page based on the Luhn score, the latent semantic analysis score, the content body score, and the supervised score. In 2 o block 901, the component selects the next sentence of the web page. In decision block 902, if all the sentences have already been selected, then the component returns the combined scores, else the component continues at block 903. In block 903, the component combines the scores for the selected sentence and then loops to block 901 to select the next sentence.
One skilled in the art will appreciate that although specific embodiments of the summarization system have been described herein for purposes of illustration, various modifications may be made without deviating from the spirit and scope of the invention. One skilled in the art will appreciate that classification refers to the process of identifying the class or category associated with a display page.
The 3 o classes may be predefined. The attributes of a display page to be classified may be compared to attributes derived from other display pages that have been classified (e.g., a training set). Based on the comparison, the display page is classified into the class whose display page attributes are similar to those of the display page being classified. Clustering, in contrast, refers to the process of identifying from a set of display pages groups of display pages that are similar to each other. Accordingly, the invention is not limited except by the appended claims.
Claims (42)
1. A method in a computer system for classifying web pages, the method comprising:
retrieving a web page;
automatically generating a summary of the retrieved web page; and determining a classification for the retrieved web page based on the automatically generated summary.
retrieving a web page;
automatically generating a summary of the retrieved web page; and determining a classification for the retrieved web page based on the automatically generated summary.
2. The method of claim 1 wherein the automatically generating of the summary includes calculating a score for each sentence of the web page using multiple summarization techniques.
3. The method of claim 2 wherein the score for each sentence is a linear combination of the scores of the multiple summarization techniques.
4. The method of claim 1 wherein the sentences with the top scores are selected to form the summary.
5. The method of claim 2 wherein the summarization techniques include a Luhn summarization technique, a latent semantic analysis summarization technique, a content body summarization technique, and a supervised summarization technique.
6. The method of claim 2 wherein the summarization techniques include any two or more of the set of summarization techniques consisting of a Luhn summarization technique, a latent semantic analysis summarization technique, a content body summarization technique, and a supervised summarization technique.
7. The method of claim 1 wherein the determining of the classification uses a Naïve Bayesian classifier.
8. The method of claim 1 wherein the determining of the classification uses a support vector machine.
9. The method of claim 1 wherein the automatically generating of a summary uses a Luhn summarization technique.
10. The method of claim 1 wherein the automatically generating of a summary uses a latent semantic analysis summarization technique.
11. The method of claim 1 wherein the automatically generating of a summary uses a content body summarization technique.
12. The method of claim 1 wherein the automatically generating of a summary uses a supervised summarization technique.
13. A method in a computer system for summarizing a web page, the method comprising:
retrieving the web page;
for each sentence of the retrieved web page, assigning a score to the sentence for each of multiple summarization techniques; and combining the scores assigned to the sentence to generate a combined score for the sentence; and selecting the sentences with the highest combined scores to form a summary of the retrieved web page.
retrieving the web page;
for each sentence of the retrieved web page, assigning a score to the sentence for each of multiple summarization techniques; and combining the scores assigned to the sentence to generate a combined score for the sentence; and selecting the sentences with the highest combined scores to form a summary of the retrieved web page.
14. The method of claim 13 wherein the combined score for each sentence is a linear combination of the assigned scores.
15. The method of claim 14 wherein the assigned scores of the multiple summarization techniques are weighted differently when combining.
16.The method of claim 13 wherein the summarization techniques include a Luhn summarization technique, a latent semantic analysis summarization technique, a content body summarization technique, and a supervised summarization technique.
17.The method of claim 13 wherein the summarization techniques include any two or more of the set of summarization techniques consisting of a Luhn summarization technique, a latent semantic analysis summarization technique, a content body summarization technique, and a supervised summarization technique.
18. The method of claim 13 wherein a summarization technique is a Luhn summarization technique in which a classification has a collection of significant words.
19. The method of claim 18 wherein noisy words are discarded from a collection.
20. The method of claim 13 wherein a summarization technique is a supervised summarization technique in which a sentence is represented by a set of features that includes a feature based on similarity between the sentence and metadata of the web page.
21.The method of claim 13 wherein a summarization technique is a supervised summarization technique in which a sentence is represented by a set of features that includes a feature based on words of the sentence that are highlighted in the web page.
22. The method of claim 13 wherein a summarization technique is a supervised summarization technique in which a sentence is represented by a set of features that includes a feature based on font size of words in the sentence.
23.The method of claim 13 including identifying a classification for the retrieved web page based on the summary of the retrieved web page.
24. The method of claim 23 wherein the identifying of the classification uses a Naive Bayesian classifier.
25. The method of claim 23 wherein the identifying of the classification uses a support vector machine.
26.A computer-readable medium containing instructions for causing a computer system to generate a summary for a display page by a method comprising:
for each sentence of the display page, generating a score for the display page that is based on multiple summarization techniques; and selecting the sentences with the highest generated scores to form a summary of the display page.
for each sentence of the display page, generating a score for the display page that is based on multiple summarization techniques; and selecting the sentences with the highest generated scores to form a summary of the display page.
27.The computer-readable medium of claim 26 wherein the generated score for each sentence is a combination of a score for each of the multiple summarization techniques.
28. The computer-readable medium of claim 27 wherein the scores of the multiple summarization techniques are weighted differently.
29.The computer-readable medium of claim 26 wherein the summarization techniques include a Luhn summarization technique, a latent semantic analysis summarization technique, a content body summarization technique, and a supervised summarization technique.
30. The computer-readable medium of claim 26 wherein the summarization techniques include any two or more of the set of summarization techniques consisting of a Luhn summarization technique, a latent semantic analysis summarization technique, a content body summarization technique, and a supervised summarization technique.
31.The computer-readable medium of claim 26 wherein a summarization technique is a Luhn summarization technique in which a classification has a collection of significant words.
32.The computer-readable medium of claim 31 wherein noisy words are discarded from the collection.
33.The computer-readable medium of claim 26 wherein a summarization technique is a supervised summarization technique in which a sentence is represented by a set of features that includes a feature based on similarity between the sentence and metadata of the display page.
34.The computer-readable medium of claim 26 wherein a summarization technique is a supervised summarization technique in which a sentence is represented by a set of features that includes a feature based on words of the sentence that are highlighted in the display page.
35.The computer-readable medium of claim 26 wherein a summarization technique is a supervised summarization technique in which a sentence is represented by a set of features that includes a feature based on font size of words in the sentence.
36.The computer-readable medium of claim 26 including identifying a classification for the display page based on the summary of the display page.
37.A computer system for classifying display pages, comprising:
means for automatically generating a summary of the display page; and means for identifying a classification for the display page based on the automatically generated summary.
means for automatically generating a summary of the display page; and means for identifying a classification for the display page based on the automatically generated summary.
38.The computer system of claim 37 wherein the means for automatically generating the summary calculates a score for each sentence of the display page using multiple summarization techniques.
39.The computer system of claim 38 wherein the score for each sentence is a linear combination of the scores of the multiple summarization techniques.
40.The computer system of claim 37 wherein the summarization techniques include a Luhn summarization technique, a latent semantic analysis summarization technique, a content body summarization technique, and a supervised summarization technique.
41.The computer system of claim 37 wherein the summarization techniques include any two or more of the set of summarization techniques consisting of a Luhn summarization technique, a latent semantic analysis summarization technique, a content body summarization technique, and a supervised summarization technique.
42.The computer system of claim 41 wherein each sentence of the display page is assigned a score that is a combination of the scores of the multiple summarization techniques.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US10/836,319 | 2004-04-30 | ||
| US10/836,319 US7392474B2 (en) | 2004-04-30 | 2004-04-30 | Method and system for classifying display pages using summaries |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CA2505957A1 true CA2505957A1 (en) | 2005-10-30 |
| CA2505957C CA2505957C (en) | 2014-10-21 |
Family
ID=34939612
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CA2505957A Expired - Fee Related CA2505957C (en) | 2004-04-30 | 2005-04-29 | Method and system for classifying display pages using summaries |
Country Status (12)
| Country | Link |
|---|---|
| US (2) | US7392474B2 (en) |
| EP (1) | EP1591924B1 (en) |
| JP (1) | JP2005322245A (en) |
| KR (1) | KR101203345B1 (en) |
| CN (1) | CN1758245B (en) |
| AT (1) | ATE470192T1 (en) |
| AU (1) | AU2005201766A1 (en) |
| BR (1) | BRPI0502155A (en) |
| CA (1) | CA2505957C (en) |
| DE (1) | DE602005021581D1 (en) |
| MX (1) | MXPA05004682A (en) |
| RU (1) | RU2377645C2 (en) |
Families Citing this family (61)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8868670B2 (en) * | 2004-04-27 | 2014-10-21 | Avaya Inc. | Method and apparatus for summarizing one or more text messages using indicative summaries |
| US7392474B2 (en) * | 2004-04-30 | 2008-06-24 | Microsoft Corporation | Method and system for classifying display pages using summaries |
| US7707265B2 (en) * | 2004-05-15 | 2010-04-27 | International Business Machines Corporation | System, method, and service for interactively presenting a summary of a web site |
| US7475067B2 (en) * | 2004-07-09 | 2009-01-06 | Aol Llc | Web page performance scoring |
| US7747618B2 (en) | 2005-09-08 | 2010-06-29 | Microsoft Corporation | Augmenting user, query, and document triplets using singular value decomposition |
| US7739254B1 (en) * | 2005-09-30 | 2010-06-15 | Google Inc. | Labeling events in historic news |
| KR100775852B1 (en) | 2006-01-18 | 2007-11-13 | 포스데이타 주식회사 | System and method for searching resource of application program |
| US20080077576A1 (en) * | 2006-09-22 | 2008-03-27 | Cuneyt Ozveren | Peer-To-Peer Collaboration |
| US7672912B2 (en) * | 2006-10-26 | 2010-03-02 | Microsoft Corporation | Classifying knowledge aging in emails using Naïve Bayes Classifier |
| US20080103849A1 (en) * | 2006-10-31 | 2008-05-01 | Forman George H | Calculating an aggregate of attribute values associated with plural cases |
| EP2080127A2 (en) * | 2006-11-01 | 2009-07-22 | Bloxx Limited | Methods and systems for web site categorisation training, categorisation and access control |
| US7617182B2 (en) * | 2007-01-08 | 2009-11-10 | Microsoft Corporation | Document clustering based on entity association rules |
| US8161369B2 (en) | 2007-03-16 | 2012-04-17 | Branchfire, Llc | System and method of providing a two-part graphic design and interactive document application |
| CN101296155B (en) * | 2007-04-23 | 2011-02-16 | 华为技术有限公司 | Contents classification method and system |
| CN101452470B (en) * | 2007-10-18 | 2012-06-06 | 广州索答信息科技有限公司 | Summary-style network search engine system and search method and uses |
| CN101184259B (en) * | 2007-11-01 | 2010-06-23 | 浙江大学 | Keyword automatically learning and updating method in rubbish short message |
| US9292601B2 (en) * | 2008-01-09 | 2016-03-22 | International Business Machines Corporation | Determining a purpose of a document |
| CN101505295B (en) * | 2008-02-04 | 2013-01-30 | 华为技术有限公司 | Method and equipment for correlating content with type |
| US8046361B2 (en) * | 2008-04-18 | 2011-10-25 | Yahoo! Inc. | System and method for classifying tags of content using a hyperlinked corpus of classified web pages |
| US20110047006A1 (en) * | 2009-08-21 | 2011-02-24 | Attenberg Joshua M | Systems, methods, and media for rating websites for safe advertising |
| JP4965623B2 (en) * | 2009-09-30 | 2012-07-04 | インターナショナル・ビジネス・マシーンズ・コーポレーション | Method, system, and program for supporting input of execution parameters of predetermined software to input field |
| JP2011095905A (en) * | 2009-10-28 | 2011-05-12 | Sony Corp | Information processing apparatus and method, and program |
| EP2482247A4 (en) * | 2009-10-30 | 2014-11-19 | Rakuten Inc | Characteristic content determination program, characteristic content determination device, characteristic content determination method, recording medium, content generation device, and related content insertion device |
| US8732017B2 (en) * | 2010-06-01 | 2014-05-20 | Integral Ad Science, Inc. | Methods, systems, and media for applying scores and ratings to web pages, web sites, and content for safe and effective online advertising |
| US9436764B2 (en) * | 2010-06-29 | 2016-09-06 | Microsoft Technology Licensing, Llc | Navigation to popular search results |
| US8635061B2 (en) | 2010-10-14 | 2014-01-21 | Microsoft Corporation | Language identification in multilingual text |
| JP5492047B2 (en) * | 2010-10-21 | 2014-05-14 | 日本電信電話株式会社 | Purchasing behavior analysis apparatus, purchasing behavior analysis method, purchasing behavior analysis program, purchasing behavior analysis system, and control method |
| US10534931B2 (en) | 2011-03-17 | 2020-01-14 | Attachmate Corporation | Systems, devices and methods for automatic detection and masking of private data |
| CN102737017B (en) * | 2011-03-31 | 2015-03-11 | 北京百度网讯科技有限公司 | Method and apparatus for extracting page theme |
| US20130066814A1 (en) * | 2011-09-12 | 2013-03-14 | Volker Bosch | System and Method for Automated Classification of Web pages and Domains |
| US9613135B2 (en) | 2011-09-23 | 2017-04-04 | Aol Advertising Inc. | Systems and methods for contextual analysis and segmentation of information objects |
| US8793252B2 (en) * | 2011-09-23 | 2014-07-29 | Aol Advertising Inc. | Systems and methods for contextual analysis and segmentation using dynamically-derived topics |
| EP2756425B1 (en) * | 2011-10-14 | 2020-11-11 | Oath Inc. | Method and apparatus for automatically summarizing the contents of electronic documents |
| US9152730B2 (en) * | 2011-11-10 | 2015-10-06 | Evernote Corporation | Extracting principal content from web pages |
| RU2491622C1 (en) * | 2012-01-25 | 2013-08-27 | Общество С Ограниченной Ответственностью "Центр Инноваций Натальи Касперской" | Method of classifying documents by categories |
| CN103324622A (en) * | 2012-03-21 | 2013-09-25 | 北京百度网讯科技有限公司 | Method and device for automatic generating of front page abstract |
| US9223861B2 (en) * | 2012-05-10 | 2015-12-29 | Yahoo! Inc. | Method and system for automatic assignment of identifiers to a graph of entities |
| US10387911B1 (en) | 2012-06-01 | 2019-08-20 | Integral Ad Science, Inc. | Systems, methods, and media for detecting suspicious activity |
| JP5700007B2 (en) * | 2012-09-13 | 2015-04-15 | キヤノンマーケティングジャパン株式会社 | Information processing apparatus, method, and program |
| US20150046562A1 (en) * | 2013-08-07 | 2015-02-12 | Convergent Development Limited | Web browser orchestration |
| CN104933055B (en) * | 2014-03-18 | 2020-01-31 | 腾讯科技（深圳）有限公司 | Webpage identification method and webpage identification device |
| US10021102B2 (en) | 2014-10-31 | 2018-07-10 | Aruba Networks, Inc. | Leak-proof classification for an application session |
| CN105786853A (en) * | 2014-12-22 | 2016-07-20 | 北京奇虎科技有限公司 | Display method and system for smart abstract of forum post |
| WO2016171709A1 (en) * | 2015-04-24 | 2016-10-27 | Hewlett-Packard Development Company, L.P. | Text restructuring |
| EP3230892A4 (en) * | 2015-04-29 | 2018-05-23 | Hewlett-Packard Development Company, L.P. | Topic identification based on functional summarization |
| RU2638015C2 (en) | 2015-06-30 | 2017-12-08 | Общество С Ограниченной Ответственностью "Яндекс" | Method for identification of target object on web page |
| US10074042B2 (en) | 2015-10-06 | 2018-09-11 | Adobe Systems Incorporated | Font recognition using text localization |
| US9875429B2 (en) | 2015-10-06 | 2018-01-23 | Adobe Systems Incorporated | Font attributes for font recognition and similarity |
| US10042880B1 (en) * | 2016-01-06 | 2018-08-07 | Amazon Technologies, Inc. | Automated identification of start-of-reading location for ebooks |
| RU2642413C2 (en) * | 2016-02-09 | 2018-01-24 | Общество С Ограниченной Ответственностью "Яндекс" | Method (versions) and server for text processing |
| US10007868B2 (en) | 2016-09-19 | 2018-06-26 | Adobe Systems Incorporated | Font replacement based on visual similarity |
| RU2635213C1 (en) * | 2016-09-26 | 2017-11-09 | Самсунг Электроникс Ко., Лтд. | Text summarizing method and device and machine-readable media used for its implementation |
| US10699062B2 (en) * | 2017-08-01 | 2020-06-30 | Samsung Electronics Co., Ltd. | Apparatus and method for providing summarized information using an artificial intelligence model |
| US10248628B2 (en) * | 2017-08-15 | 2019-04-02 | Hybris Ag | Statistical approach for testing multiple versions of websites |
| US10579698B2 (en) | 2017-08-31 | 2020-03-03 | International Business Machines Corporation | Optimizing web pages by minimizing the amount of redundant information |
| EP3966999A4 (en) * | 2018-10-10 | 2023-07-05 | Scaramanga Technologies Pvt. Ltd. | Method, system and apparatus for providing a contextual keyword collective for communication events in a multi-communication platform environment |
| US11397776B2 (en) | 2019-01-31 | 2022-07-26 | At&T Intellectual Property I, L.P. | Systems and methods for automated information retrieval |
| US10950017B2 (en) | 2019-07-08 | 2021-03-16 | Adobe Inc. | Glyph weight modification |
| US11295181B2 (en) | 2019-10-17 | 2022-04-05 | Adobe Inc. | Preserving document design using font synthesis |
| CN111797945B (en) * | 2020-08-21 | 2020-12-15 | 成都数联铭品科技有限公司 | Text classification method |
| US20230222149A1 (en) * | 2022-01-11 | 2023-07-13 | Intuit Inc. | Embedding performance optimization through use of a summary model |
Family Cites Families (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH02254566A (en) * | 1989-03-29 | 1990-10-15 | Nippon Telegr & Teleph Corp <Ntt> | Automatic excerpt generating device |
| US5317507A (en) | 1990-11-07 | 1994-05-31 | Gallant Stephen I | Method for document retrieval and for word sense disambiguation using neural networks |
| JP2944346B2 (en) * | 1993-01-20 | 1999-09-06 | シャープ株式会社 | Document summarization device |
| US5918240A (en) * | 1995-06-28 | 1999-06-29 | Xerox Corporation | Automatic method of extracting summarization using feature probabilities |
| US5864855A (en) | 1996-02-26 | 1999-01-26 | The United States Of America As Represented By The Secretary Of The Army | Parallel document clustering process |
| JPH09319768A (en) * | 1996-05-29 | 1997-12-12 | Oki Electric Ind Co Ltd | Main point extracting method |
| US6359633B1 (en) * | 1999-01-15 | 2002-03-19 | Yahoo! Inc. | Apparatus and method for abstracting markup language documents |
| NZ515293A (en) * | 1999-05-05 | 2004-04-30 | West Publishing Company D | Document-classification system, method and software |
| JP4320491B2 (en) * | 1999-11-18 | 2009-08-26 | ソニー株式会社 | Document processing system, terminal device, document providing device, document processing method, recording medium |
| US7137065B1 (en) * | 2000-02-24 | 2006-11-14 | International Business Machines Corporation | System and method for classifying electronically posted documents |
| US6606644B1 (en) | 2000-02-24 | 2003-08-12 | International Business Machines Corporation | System and technique for dynamic information gathering and targeted advertising in a web based model using a live information selection and analysis tool |
| US6775677B1 (en) * | 2000-03-02 | 2004-08-10 | International Business Machines Corporation | System, method, and program product for identifying and describing topics in a collection of electronic documents |
| US7130848B2 (en) * | 2000-08-09 | 2006-10-31 | Gary Martin Oosta | Methods for document indexing and analysis |
| US7607083B2 (en) * | 2000-12-12 | 2009-10-20 | Nec Corporation | Test summarization using relevance measures and latent semantic analysis |
| US20020087326A1 (en) * | 2000-12-29 | 2002-07-04 | Lee Victor Wai Leung | Computer-implemented web page summarization method and system |
| US20020169770A1 (en) * | 2001-04-27 | 2002-11-14 | Kim Brian Seong-Gon | Apparatus and method that categorize a collection of documents into a hierarchy of categories that are defined by the collection of documents |
| US6609124B2 (en) | 2001-08-13 | 2003-08-19 | International Business Machines Corporation | Hub for strategic intelligence |
| US20040205457A1 (en) * | 2001-10-31 | 2004-10-14 | International Business Machines Corporation | Automatically summarising topics in a collection of electronic documents |
| US20030221163A1 (en) * | 2002-02-22 | 2003-11-27 | Nec Laboratories America, Inc. | Using web structure for classifying and describing web pages |
| US6910037B2 (en) * | 2002-03-07 | 2005-06-21 | Koninklijke Philips Electronics N.V. | Method and apparatus for providing search results in response to an information search request |
| US7130837B2 (en) * | 2002-03-22 | 2006-10-31 | Xerox Corporation | Systems and methods for determining the topic structure of a portion of text |
| US7065707B2 (en) * | 2002-06-24 | 2006-06-20 | Microsoft Corporation | Segmenting and indexing web pages using function-based object models |
| US7292972B2 (en) * | 2003-01-30 | 2007-11-06 | Hewlett-Packard Development Company, L.P. | System and method for combining text summarizations |
| US7392474B2 (en) | 2004-04-30 | 2008-06-24 | Microsoft Corporation | Method and system for classifying display pages using summaries |
-
2004
- 2004-04-30 US US10/836,319 patent/US7392474B2/en not_active Expired - Fee Related
-
2005
- 2005-04-27 AU AU2005201766A patent/AU2005201766A1/en not_active Abandoned
- 2005-04-28 BR BR0502155-3A patent/BRPI0502155A/en not_active IP Right Cessation
- 2005-04-29 DE DE602005021581T patent/DE602005021581D1/en active Active
- 2005-04-29 CA CA2505957A patent/CA2505957C/en not_active Expired - Fee Related
- 2005-04-29 RU RU2005113190/09A patent/RU2377645C2/en not_active IP Right Cessation
- 2005-04-29 MX MXPA05004682A patent/MXPA05004682A/en not_active Application Discontinuation
- 2005-04-29 AT AT05103580T patent/ATE470192T1/en not_active IP Right Cessation
- 2005-04-29 EP EP05103580A patent/EP1591924B1/en not_active Not-in-force
- 2005-04-29 KR KR1020050036077A patent/KR101203345B1/en active IP Right Grant
- 2005-04-30 CN CN2005100896481A patent/CN1758245B/en not_active Expired - Fee Related
- 2005-05-02 JP JP2005134491A patent/JP2005322245A/en active Pending
-
2008
- 2008-06-24 US US12/145,222 patent/US20090119284A1/en not_active Abandoned
Also Published As
| Publication number | Publication date |
|---|---|
| CN1758245B (en) | 2010-09-08 |
| US7392474B2 (en) | 2008-06-24 |
| AU2005201766A1 (en) | 2005-11-17 |
| CN1758245A (en) | 2006-04-12 |
| KR101203345B1 (en) | 2012-11-20 |
| MXPA05004682A (en) | 2005-11-17 |
| RU2005113190A (en) | 2006-11-10 |
| KR20060047636A (en) | 2006-05-18 |
| EP1591924B1 (en) | 2010-06-02 |
| ATE470192T1 (en) | 2010-06-15 |
| JP2005322245A (en) | 2005-11-17 |
| US20090119284A1 (en) | 2009-05-07 |
| US20050246410A1 (en) | 2005-11-03 |
| CA2505957C (en) | 2014-10-21 |
| BRPI0502155A (en) | 2006-01-10 |
| RU2377645C2 (en) | 2009-12-27 |
| EP1591924A1 (en) | 2005-11-02 |
| DE602005021581D1 (en) | 2010-07-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CA2505957C (en) | Method and system for classifying display pages using summaries | |
| Munot et al. | Comparative study of text summarization methods | |
| Delort et al. | Enhanced web document summarization using hyperlinks | |
| Ceri et al. | Web information retrieval | |
| Agrawal et al. | A detailed study on text mining techniques | |
| Liu et al. | Special issue on web content mining | |
| US8645370B2 (en) | Scoring relevance of a document based on image text | |
| US8156097B2 (en) | Two stage search | |
| US7363279B2 (en) | Method and system for calculating importance of a block within a display page | |
| Macdonald et al. | Blog track research at TREC | |
| EP2122506A1 (en) | Method and system for information discovery and text analysis | |
| Paranjpe | Learning document aboutness from implicit user feedback and document structure | |
| Braslavski | Marrying relevance and genre rankings: an exploratory study | |
| Stamou et al. | Classifying web data in directory structures | |
| Han et al. | Joint web-feature (JFEAT): a novel web page classification framework | |
| Liu et al. | Algorithm for ranking news | |
| Liu et al. | Keyphrase extraction for labeling a website topic hierarchy | |
| Leong et al. | Text summarization for knowledge filtering agents in distributed heterogeneous environments | |
| Patil et al. | Optimized summarization of research papers as an aid for research scholars using data mining techniques | |
| Pun et al. | Ranking Search Results by Web Quality Dimensions. | |
| Syed et al. | User preference modeling-a survey report | |
| Kumaar et al. | Effective Online Discussion Data for Teachers Reflective Thinking Using Feature base Model | |
| Qumsiyeh et al. | Web Search Using Summarization on Clustered Web Documents Retrieved by User Queries | |
| Toh | Design of a Low‐Cost Adaptive Question Answering System for Closed Domain Factoid Queries | |
| Trajkovski | Computer Generated News Site–TIME. mk |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| EEER | Examination request | ||
| MKLA | Lapsed |
Effective date: 20220301 |
|
| MKLA | Lapsed |
Effective date: 20200831 |