FIELD OF THE INVENTION
The invention relates in particular to a method for categorizing web sites that provide audio broadcasts over the Internet.
Currently, nearly 10,000 radio stations broadcast over the Internet. The stations stream their audio content. A streamed file is a file that can be started for playing out before the download is completed. With a proper network connection, decoding and playback software on one's PC or set top box the audio can be captured. Audio output hardware, e.g., analog sound cards, USB speakers, and streaming media tools, such as RealPlayer from RealNetworks, Inc., have become widely available and enable to add the radio functionality to one's PC.
SUMMARY OF THE INVENTION
The user has to check many listings of large numbers of web sites that provide these broadcasts over the Internet. Accordingly, there is a need for helping the user to select from among the huge number of stations available. To this end, the invention provides a method for categorizing web sites or resources on the Internet that provide audio (e.g., speech and music) streaming based on their typical content. A web resource that provides audio streaming is identified by its resource type. The resource type is determined by way of the type extension in its URL that indicates the file format, e.g., “.ram”, “.tsp” or “.swa”. This extension enables, for example, to automatically open the proper software applications (or “plug-ins”) in the user's browser when the hyperlink is clicked. Accordingly, the relevant resources on the Internet can be identified based on their URL. If the file extension is not available through the URL, the resource type is determined by the MIME type or content-type information provided in the HTTP header of the resource. Taking into consideration the resource's country domain extension, e.g., “.nl” for the Netherlands or “.ru” for Russia, further optimizes the analysis of the URL, for example if one is interested in audio content in a specific natural language. Upon finding a relevant resource, i.e., one that provides streaming of audio, the resource's file is retrieved from the relevant server and analyzed based on its audio content. In a preferred embodiment, the inventor proposes to use speech recognition or music (tune/rhythm) recognition software to search through and categorize these stations by, e.g., language, style of music, absence of commercials. Speech recognition software is capable of determining the signature of various kinds of music, thus allowing categorization of music with just this kind of software. For example, classical music has typically a different speech recognition signature than rock music. A server can be dedicated to categorize stations or channels in a data base, similar as to what PlanetSearch or Altavista does for text documents. One or more web crawlers can be used in parallel to automatically fetch web sites that supply audio so as to identify them for a search engine. Additionally, the resource's server can be evaluated by the crawler for the quality of the connection, e.g., connection speed, reliability, etc. For example, the categorizing server may recommend to a user, who has broadband network access (e.g., ISDB, cable, T1), higher connection speed sources. An audio browser is provided, analogous to PlanetSearch's or Alta Vista's for text, to provide a searchable collection of Internet audio web sites based from which specific pages are returned to the user based on certain audio search criteria. Alternatively, the catalog approach (Yahoo experts hand-pick and assign sites to categories) can be taken to categorize the stations at the server and make them accessible through a search engine. Once the sites are categorized, a user provides a query input to the server and receives a list of URLs representative of the channels that match the query input (e.g., give me a French language station that plays music like this). As an alternative or supporting this, the server provides a customized electronic program guide to the user based on a profile of the user stored on the server, e.g., using the SmartConnect infrastructure of Philips Electronics.
The invention is of commercial interest in particular to, e.g., cable providers and network owners, and service providers in order to serve as an incentive for subscribers.
As to music recognition, see, for example U.S. Ser. No. 08/840,356 (Attorney Docket PHA 23,241) filed Apr. 28, 1997 for Mark Hoffberg for BIBLIOGRAPHIC MUSIC DATA BASE WITH NORMALIZED MUSICAL THEMES, now allowed and herein incorporated by reference. This patent document discusses, among other things, how rhythm information or tonal information of a musical theme can be used to identify the theme. The rhythm information comprises the time signature (meter) and the accentuations of the theme. The time signature determines the number of beats to the measure. The accentuation determines which beat gets an accent and which one does not. For example, the sign 6 8 in a musical score is the time signature indicating that the meter is 6 beats to the measure and that an eighth note gets one beat. Flamenco music has a variety of different styles, each determined by its own compŗs (rhythmic accentuation pattern). Typical examples of flamenco music are Alegrias, Buleržas, Siguiriyas and Soleares that all have 12 beats to the measure. In the Alegrias, Buleržas and Soleares, the third, sixth, eighth, tenth and twelfth beats are accentuated. The first, third, fifth, eighth and eleventh beats are emphasized in the Siguiriyas style. In this system rhythmic accentuation patterns are used as input data in order to retrieve bibliographic information associated with the theme that is represented by the rhythm. For example, the rhythmic accentuation pattern is entered into the system as a substantially monotonic sequence of accentuated and unaccentuated sounds. The input data then is represented by, e.g., a sequence of beats or peaks of varying height in the time domain. The relative distances between successive peaks represent the temporal aspects of the pattern and the relative heights represent the accentuations in the pattern. The sequence of beats and rests in between is represented by a digital word. The words can be stored lexicographically to enable a fast and orderly retrieval. If tonal information and/or rhythm information can be used to identify individual musical themes, they can also be used to identify with more or less accuracy a certain style of music.
As to SmartConnect, see, for example U.S. Ser. No. 09/160,490 (attorney docket PHA 23,500) filed Sep. 25, 1998 for Adrian Turner et al., for CUSTOMIZED UPGRADING OF INTERNET-ENABLED DEVICES BASED ON USER-PROFILE, herein incorporated by reference. This document relates to a server system that maintains a user profile of a particular end-user of consumer electronics network-enabled equipment and a data base of new technical features for this type of equipment, e.g., a home network. If there is a match between the user-profile and a new technical feature, and the user has indicates to prefer receiving information about updates or sales offers, the user gets notified via the network of the option to obtain the feature.
As to SmartConnect, also see U.S. Ser. No. 09/189,535 (attorney docket PHA 23,527) filed Nov. 10, 1998 for Yevgeniy Shteyn for UPGRADING OF SYNERGETIC ASPECTS OF HOME NETWORKS, herein incorporated by reference. This document relates to a system with a server that has access to an inventory of devices and capabilities on a user's home network. The inventory is, for example, a look-up service as provided by HAVi, JINI and Home API architectures. The server has also access to a data base with information of features for a network. The server determines if the synergy of the apparatus present on the user's network can be enhanced based on the listing of the inventory and on the user's profile. If there are features that are relevant to the synergy, based on these criteria, the user gets notified.
FIG. 2 illustrates this in more detail with reference to a block diagram of a system 200. System 200 comprises a dedicated server 202 that sends out multiple spiders 204, 206, . . . , 208 over the Internet 210 to visit Web sites. As known, a spider, also referred to as a “web crawler”, is a software program that fetches Web pages and analyzes their content in order to generate searchable, indexed catalogs for a search engine. Web sites and specific pages can be visited and indexed selectively. A typical Web page includes one or more hyperlinks to other Web pages. Therefore, a spider can start almost from anywhere and hop from Web page to Web page following the links it encounters while being out there on the Internet. Each of spiders 204-208 visits HTML pages and scans it for clickable links that indicate the presence of resources for streamable audio. In this sense, an audio spider, i.e., the entity that specifically looks for audio links, may ride piggyback with a conventional spider or crawler that scans text-based information. Currently, popular formats for streaming audio include RealAudio (file extension “.ram”) from RealNetworks, Inc., TrueSpeech (file extension “.tsp”) from DSP Group, Inc., and Macromedia's Shockwave for Director (file extension “.swa”). Links that have these extensions are relevant to spiders 204-208.
Once a spider, e.g., spider 206, has identified a resource 212 (based on its hyperlink 214) that provides audio streaming, it fetches the data and causes the content to be analyzed, e.g., by server 202, based on the content's pattern. The content's pattern is analyzed using, for example, automated speech recognition methods 216, or automated music pattern or rhythmic pattern analysis 218 as discussed above. Based on the results of this analysis, the content is indexed by an indexer 220 in a data base 222 as relating to a certain natural language, one or more music styles. Alternatively, or subsidiarily, human experts 224 listen to the content associated by the links thus identified by the spiders and categorize the audio by hand in data base 222. As a side remark, note that the number of audio sites on the Internet is large but not as nearly as large as the number of textual and/or graphics sites by many orders of magnitude. Accordingly, it pays off to have the audio links scanned or data base 222 reviewed by human experts. Server 202 provides a search engine 226 to search data base 222 for specific concepts requested by a user via his/her client 228 that has a browser 230. For example, the user requests audio sites that supply a live radio broadcast in the Spanish language. The user submits his/her request that has the terms “live” and “Spanish” (or the Spanish equivalent thereof) in it. Sites that provide Spanish spoken programs are recognized by their language, e.g., via speech recognition 216. Sites that provide a live broadcast are recognized as such by experts 224, or may be automatically identified by listening in on the repeated mentioning throughout the day and at certain regular time intervals of the date and time, as is the case with most news services. Alternatively or subsidiarily, the meta tags associated with the HTML pages containing the audio hyperlinks or associated with the audio pages themselves contain the expression “live” (or the equivalent in another language). The cross section of the set of audio sites that provide “live” broadcasts with the set of audio sites that provide audio comprising Spanish language is the set wherein the requester is interested. Similarly, if the user is interested in audio streaming of a certain piece of music, he or she may provide the input in a format as disclosed in U.S. Ser. No. 08/840,356 (PHA 23,241) of the same inventor, now allowed and discussed above, to determine if there is a matching resource available on the Internet. If the user is interested in sites that supply a certain style of music, he or she may submit a request to search engine 226 in a textual format, such as: “get me the sites that provide French chansons from the fifties”. The relevant terms here are: “French”, “chansons” and “fifties” based on which data base 222 is queried. Whether or not a certain chanson is from the time period indicated could have been added as an entry to data base 222 by experts 224. Alternatively, or as an additional support the music data base discussed in U.S. Ser. No. 08/840,356 (PHA 23,241) is used to intervene and to convert the user's request into a request for specific music titles queried in the manner specified.