FR2998998A1 - Method for obtaining set of metadata associated with document for research in newsgroup, involves selecting document, and predetermining weighting value according to opinion user relative to response to preliminary request - Google Patents

Abstract

The method involves transmitting a current request comprising a keyword, and receiving a response to the current request containing a set of metadata (MD1-MDn) associated with a document (D1). The document is selected according to a relevance value given for the document according to a weighting value associated with the keyword of the current request that is in turn associated with the document. The weighting value is predetermined according to an opinion user relative to the response to a request that is preliminary to the current request. Independent claims are also included for the following: (1) a method for transmission of a set of metadata (2) a terminal (3) a server.

Description

The invention relates to the field of retrieval of documents in one or more document databases. Known techniques make it possible to search documents in document databases. These known techniques are based on search engines which, based on key words entered for example by a user, search in one or more databases, documents containing the selected keywords. These systems comprise, in addition to a search engine, a memory area in which is recorded a dictionary of keywords, generally called vocabulary or corpus. This vocabulary is pre-established. A weighting value is associated with each word of the vocabulary.

When searching for a document, following the entry of keywords, the search engine determines whether the keywords entered are present in the vocabulary and performs a search from the key words entered in the vocabulary. These systems are particularly suitable for searching documents in a particular technical area well circumscribed. Indeed, the keywords associated with a technical field are known and vary little. Moreover, the weight of these words relative to each other does not vary over time. The weighting values are thus defined only once. As a result, vocabulary and weighting values remain appropriate even when documents are added or removed from the database. These systems are therefore not well suited to finding documents in databases containing documents relating to one or more topics whose boundaries are poorly defined, for example for research in a discussion forum. In this case, the results obtained are not sufficiently relevant. There is therefore a need to improve the results provided in response to a request for documents from a database comprising documents dealing with a plurality of topics. The invention improves the situation. For this purpose, the invention relates to a method for obtaining at least one set of metadata associated with at least one document selected from among a plurality of documents, characterized in that it comprises the following steps: a current request containing at least one keyword; -2- - receiving a response to said current request containing at least one set of metadata associated with at least one document, said at least one document being selected from the documents of the plurality of documents based on at least one value of determined relevance for said at least one document based on at least one weighting value associated with at least one keyword of said current request associated with said document, said weighting value being predetermined according to at least one relative user opinion. at least one response to at least a first request prior to said current request. User reviews made by users after receiving responses to previous queries can determine one or more weight values for one or more keywords. The user's notice expresses the interest that a user has found in a response provided for a request that he has made and therefore the relevance of one or more documents cited in this response. Information on the relevance of one or more documents is used to assign a weighting value to the keywords of the documents. Thanks to the weighting values thus determined, a response to a current request made by a user is more relevant. It corresponds more to the expectations of the user. According to a particular embodiment, the obtaining method comprises a step of displaying metadata extracted from said at least one set received followed by a step of transmitting a user notification relating to at least one document targeted in said response. The transmitted user notification contributes to the subsequent update of weighting values. According to a particular embodiment of the method of obtaining, said metadata comprise a pointer allowing access to a document and wherein the method comprises a step of displaying said pointer, a step of accessing said document via said pointer, a step of receiving a request for user notification for said document and a step of transmitting a user opinion on the relevance of said document in response to said request for user notification.

The receipt of a request for an opinion following a user's access to a document allows the user to formulate an opinion on the relevance of this document for the current request. Sending a user notice in response to a request for an opinion is a simple way for the user to express his opinion. The invention also relates to a method for transmitting at least one set of metadata associated with at least one document selected from a plurality of documents. According to the invention, the transmission method comprises the following steps: receiving a current request comprising at least one keyword; selecting among the documents from the plurality of documents at least one document based on at least one value of relevance determined for the at least one document based on at least one weighting value associated with at least one keyword said current request associated with said document, said weighting value being predetermined according to at least one user notification relating to at least one response to at least a first request prior to said current request; transmitting a response to said current request containing at least one set of metadata relating to at least one selected document.

The weighting values assigned to keywords based on user notifications received in advance allow you to select relevant documents during common queries and thus to obtain answers to requests that are satisfactory to the user, regardless of the type of documents included in the request. plurality of documents. According to a particular characteristic of the transmission method, said relevance value is furthermore determined as a function of an occurrence number of said at least one keyword of the request associated with said at least one document. The occurrence number of a keyword associated with a document is indicative of the importance of the keyword for that document. Taking this characteristic into account in determining the weighting value associated with the keyword makes it possible to present the user with answers to requests adapted to his search. According to a particular embodiment, the transmission method comprises the following steps, performed prior to said step of receiving a current request: reception steps from a plurality of terminals, first requests containing at least one keyword ; steps of transmitting responses to the first requests, respectively comprising at least one set of metadata associated with at least one document selected from the plurality of documents as a function of at least one keyword of a first request; steps of receiving user notifications for at least one of the transmitted responses; a step of determining at least one weighting value of at least one keyword based on at least one received user notification. The collection of user opinions on responses to first queries makes it possible to determine keyword weighting values. The weighting values are then used to select documents that are adapted to the users' expectations. The collection of notices from a plurality of users makes it possible to provide weighting values adapted to these users. These steps performed during a learning phase make it possible to initialize the system. According to one embodiment of the transmission method, the user's opinion relates to the relevance of a targeted document in a response to a first request and the step of receiving the notification is followed by a step determining at least one order relationship between said target document and another targeted document in said response and said at least one weighting value of at least one keyword is determined based on at least one relationship of determined order. In this embodiment, a user's notice is translated as one or more order relationships between the documents. Such a relation of order makes it possible to define for two documents which is the most relevant of the two. This information is then used to determine weighting values of keywords contained in the query. According to a particular embodiment of the transmission method, the step of transmitting a response to a current request is followed by the following steps: detection of a user action relating to a targeted document in the response; - transmitting a request for user notification following the said detection of a user action; - receipt of a user notice relating to said document; determining at least one order relationship between at least one keyword of the current request associated with said document and at least one keyword of the current request associated with another targeted document in said response; recording of said at least one determined order relationship In this embodiment, a user's notice is translated as one or more order relationships between keywords of the request. This information is then used to determine keyword weighting values. According to a particular embodiment, the transmission method comprises a step of updating at least one weighting value of at least one keyword according to at least one registered order relationship. The order relationships determined as users' opinions are gathered, that is, as requests are received, are then used to update the weighting values. The update makes it possible to take into account the evolution over time of the relevance of the documents, and consequently of the evolution in the time of the relevance of the key words. The invention also relates to a terminal characterized in that it comprises: a transmission module of a current request comprising at least one keyword; a module for receiving a response to said current request containing at least one set of metadata associated with at least one document, said at least one document being selected from documents of a plurality of documents as a function of at least one value of relevance determined for said at least one document according to at least one weighting value associated with at least one keyword of said current request associated with said document, said weighting value being predetermined according to at least one a user notification relating to at least one response to at least a first request prior to said current request.

The invention also relates to a server characterized in that it comprises: a module for receiving a current request comprising at least one keyword; a selection module among documents of a plurality of documents of at least one document as a function of at least one determined value of relevance for the said at least one document as a function of at least one weighting value associated with the document; at least one keyword of said current request associated with said document, said weighting value being predetermined as a function of at least one user notification relating to at least one response to at least a first request prior to said current request; a module for transmitting a response to said current request containing at least one set of metadata relating to at least one selected document.

The invention also relates to a computer program product comprising instructions for implementing the steps of the obtaining method as described above, when it is loaded and executed by a processor. The invention finally relates to a computer program product comprising instructions for implementing the steps of the transmission method as described above, when it is loaded and executed by a processor. Other features and advantages of the present invention will appear in the following description of embodiments given as non-limiting examples, with reference to the accompanying drawings, in which: - Figure 1 is a diagram illustrating a system according to a first embodiment of the invention; FIG. 2 is a flowchart illustrating the various steps of a method for obtaining, respectively a transmission method, according to a first embodiment; FIG. 3 is a diagram illustrating a system according to a second embodiment of the invention, - Figures 4a and 4b respectively represent a first and a second part of a flowchart illustrating the different steps of a method for obtaining respectively a transmission method according to a second embodiment, FIG. 5 is a block diagram representing a server implementing a transmission method according to a Embodiment of the invention, FIG. 6 is a block diagram showing a terminal implementing a method of obtaining according to one embodiment of the invention. A first embodiment of a method for obtaining, respectively a method for transmitting at least one set of metadata associated with at least one document selected from a plurality of documents, will now be described with reference to FIGS. and 2. With reference to FIG. 1, a system SY1 comprises a terminal T of a user U and a server S accessible by the terminal T via a communication network R.

The SY1 system also includes a database BD accessible by the server S and hosting a plurality of documents. As an alternative, the SY1 system comprises several databases BD each hosting a plurality of documents. The database (s) BD are accessible by the server S by a direct link, for example a wired or wireless link, or via the communication network R. The database or databases host N documents D1, D2 ... di ... Dn. A document Di recorded in the database BD is for example an audio and / or video content, a web page, a text file, a file containing text, one or more images, drawings, photographs, etc.

The system SY1 also includes a memory space MD, accessible by the server S, in which are recorded sets of metadata MD1, MD2 ... MDi ... MDn defined respectively for the documents D1, D2 ... Di ... Dn. For example, a set of metadata MDi is registered in association with a reference of the document Di with which it is associated.

A set of metadata MDi associated with a document Di comprises, for example, data describing the document Di, for example a summary of the content of the document, one or more authors, a date, and / or data making it possible to access the document. . The data enabling access to a document are for example an address or a pointer to the document, for example a URL (for `Uniform Resource Location ').

One or more keywords are also defined for each document and recorded in association with the metadata associated with the document considered in the memory space MD. Alternatively, the keywords defined for each document are recorded separately from the sets of metadata, for example in another memory accessible by the server S. Alternatively, each set of metadata MDi is stored in the database. BD data in association with the document Di with which it is associated. -7- A defined keyword for a document represents a keyword associated with this document. The SY1 system also includes a memory space accessible by the server S, in which a TAB key word table is stored. The table TAB groups key words Ml, M2 ... Mi ... Mz extracted from the documents D1, D2 ... Di ... Dn. Each Mi keyword is a keyword defined for at least one of the plurality of documents. The steps of a reception method and a method for transmitting at least one set of metadata associated with at least one document selected from among a plurality of documents implemented in the SY1 system are illustrated with reference to FIG. 2 The steps E2 and E12 represent the steps of the method of obtaining implemented for example by the terminal T. The steps E4, E6, E8 and El0 represent the steps of the transmission method implemented for example by the server S.

In a previous step EO, a weighting value Pi is determined for each keyword Mi of the table of keywords TAB and recorded in association with this keyword in the table of keywords TAB. The weighting value Pi is determined according to at least one user notification relating to at least one response to at least a first request. An embodiment of the determination of weighting values is described in the following description. During a step E2, the user U of the terminal T formulates a current request RQ1 containing at least one keyword. The request RQ1 is for example entered by means of an interface of a browser installed on the terminal T.

The request RQ1 is written in natural language. It includes one or more sentences, for example 'I am looking for documents on Zidane'. As an alternative, the current request RQ1 contains for example a sequence of one or more keywords, for example "Zidane". The request RQ1 is then transmitted to the server S via the telecommunications network R. The current request RQ1 is received by the server S during a step E4. In a step E6, a value of relevance VPi is determined for each document Di of the plurality. The relevance value VPi of the current request RQ1 for a document Di stored in the database BD is computed as a function of at least one weighting value Pj associated with at least one key word Mj extracted from the current request RQ1 and - 8- associated with the document Di and depending on the number of occurrence of the keyword or keywords for the document Di. The number of occurrences of each keyword for the document Di is for example predefined during a previous step and recorded in association with the keyword considered.

The relevance value VPi of a document Di for the current request RQ1 is for example determined by the equation: VPi = Pj ln (1 + (e-1) nbr (Mj) nbd (MD) MjERQ1nDi in which Mj is a key word, Pj the weighting value associated with the keyword Mj in the table of key words TAB, ln is the natural logarithm, e is the constant of Neper (approximately equal to 2.7), nbr (Mj) is the number of occurrences of the word Mj in the current request RQ1 and nbd (Mj) is the number of occurrences of the word Mj for the document Di. As an alternative, the value of relevance VPi determined for a document Di recorded in a database of data BD is calculated based on at least one weighting value associated with at least one keyword of the current request RQ1 associated with the document Di and does not take into account the number of occurrences.

In the case where the request RQ1 is formulated in natural language, key words are extracted beforehand from the content of the request RQ1. A keyword is a word extracted from the request RQ1 and appearing in the table of the keywords TAB. During a step E8, documents Dk are selected from the plurality of documents as a function of the relevance values calculated during step E6.

For example, the documents Da, Db, Dc, Dd and De for which the relevance value is highest for the current request RQ1 considered are selected. Alternatively, only documents Dk for which the relevance value is greater than a predefined threshold are selected. More generally, the documents are selected according to one or more predefined criteria relating to the relevance values of the documents of the plurality of documents determined for the current request RQ1. Then, during a step E10, a response RP1 to the current request RQ1 is transmitted by the server S to the terminal T. The response RP1 contains for each document Dk selected in step E8, all or part of the set of MDk metadata associated with the document Dk. The response RP1 is received by the terminal T and displayed on a screen of the terminal T during a step E12. A second embodiment of a method for obtaining, respectively a method for transmitting at least one set of metadata associated with at least one document selected from among a plurality of documents, will now be described with reference Figures 3, 4a and 4b. With reference to FIG. 3, a system SY2 comprises a plurality of terminals T1, T2, T3... Tk and a server S accessible by the terminals Ti, T2, T3... Tk via an communication network R. The system SY2 also comprises a database BD accessible by the server S and hosting a plurality of documents D1, D2 ... Di ... Dn. A document Di recorded in the database BD is for example an audio and / or video content, a web page, a text file ... The document Di may also include one or more images, drawings, photographs ... A description is associated with each document. The description is textual content describing the contents of the document. The description of a document is recorded in association with this document in the database BD. The system SY2 also comprises a memory space MD, accessible by the server S, in which are recorded sets of metadata MD1, MD2 ... MDi ... MDn respectively defined for the documents D1, D2 ... Di ... Dn. The MDi set of metadata is for example registered in association with a reference of the document Di with which it is associated. The set of metadata MDi associated with a document Di includes data enabling access to the document, for example an address or a pointer to the document, for example a URL (for `Uniform Resource Location '). As an alternative, a set of metadata is associated with several documents. As an alternative, the description associated with a document is recorded in association with the document metadata in the memory space MD.

Also, as an alternative, the set of metadata also includes data describing the document Di, for example a summary of the content of the document, one or more authors, a date ... The system 5Y2 also includes a memory space, accessible by the server S, in which is recorded a table of keywords TAB.

The steps of a reception method and a method for transmitting at least one set of metadata associated with at least one document selected from a plurality of documents, implemented in the system 5Y2, are illustrated with reference to the Figure 4a and 4b. In a first step E20 (FIG. 4a), one or more key words Mij are determined for each document Di of the plurality of documents. The determination of the keywords Mij comprises, for example, a step of extracting the words of the description associated with the document Di considered and then a step of filtering unwanted words. The unwanted words -10- are for example the short words such as articles, punctuations ... Moreover, the words appearing several times in the description associated with the document Di are represented only once. The extraction of words is for example carried out according to a method described in the document entitled "Foundations of Statistical Natural Language Processing" by Christophe D. Manning and Hinrich Schütze, published by "MIT Press". The one or more keywords Mij determined for each document Di are then recorded in a memory accessible by the server S in association with an identifier of the document Di.

An occurrence number nbd (Mij) of each keyword Mij in the description associated with the document Di is determined and recorded in association with the keyword considered. The occurrence number nbd (Mij) of a keyword Mij of the document D1 represents an occurrence number of the keyword Mij for the document Di. During a step E22, a table of keywords TAB is determined and stored in a memory accessible by the server S. The table of keywords TAB comprises key words Ml, M2 ... Mi ... Mz. The keyword table TAB contains the determined keywords for each document of the plurality of documents in step E20. Keywords represented in multiple Di documents are shown only once in the TAB keyword table.

During a step E24, a weighting value Pi associated with each keyword Mi of the table of keywords is initialized to a predefined value, for example the value 1, then it is recorded in the table of the keywords TAB in association with the Mi keyword it is associated with. During a step E26, a user request RQ containing one or more keywords is transmitted by one of the terminals T1, T2 ... Tk to the server S. The request RQ represents a first user request. The request RQ is received by the server S during a step E28. In a step E30, the server S selects one or more documents among the documents of the plurality, for example D5, D10 and D77. The selection includes for example the calculation of a value of relevance of each document of the plurality, for the RQ request and the selection of the documents having the highest value of relevance. The relevance value of a document is determined based on the presence or absence of one or more keywords in the RQ query in the document and the weighting value associated with the keywords in the word table. TAB keys.

The relevance value VPi of a document Di for the current request RQ is for example determined by the equation: VPi = Pj 16 (1 + (e-1) nbr (MDribd (MD) MjERQnDi in which Mj is a key word , Pj the weighting value associated with the keyword Mj, ln is the natural logarithm, e is the Neper constant (approximately equal to 2.7), nbr (Mj) is the number of occurrences of the word Mj in the current request RQ and nbd (Mj) is the number of occurrences of the word Mj in the description associated with the document Di.

As an alternative, the relevance value does not take into account the weighting value of the keywords. For example, the relevance value of a document is determined based on the occurrence number of each keyword in the description associated with the document. An answer RP to the request RQ is transmitted, during a step E32 via the telecommunications network R by the server S to the terminal having transmitted the request RQ. The response RP contains for example, for each selected document, a URL for storing the document in the database BD. The URL represents a set of metadata associated with a document, determined during a preliminary step and stored in a memory accessible by the server S. The response RP comprises for example the URLs URL5, URL10 and URL77 respectively to access selected documents D5, D10 and D77. During a step E34, the response RP is received by the terminal having transmitted the request RQ then the contents of the response RP, that is to say the URLs, is displayed on the screen of this terminal. During a step E36, a user of the terminal considered selects a URL address from the URLs displayed on the screen, for example the URL10. For example, the selection is made in a known manner by clicking on the screen of the terminal at a position on the screen where the URL is displayed by means of a computer mouse. This selection represents an AC user action vis-à-vis the associated document. In known manner, the "click" on the URL address10 causes the terminal to send an access request to the document D10 associated with the selected URL. Document D10 is received in response and displayed on the terminal screen. During a step E38, the server S detects the user action AC, that is to say the selection of a document by the user, and sends to the terminal in question, a user request request DA relative to the selected document D10 during the AC user action. The user request request DA is for example a text message asking the user if the document D10 is relevant to its request RQ. During a step E40, the terminal sends a response RA to the request for notification DA. The answer contains a user-generated AV notice. The response is received by the server S during a step E42. In a step E44, the server S analyzes the received response RA. More specifically, the server S analyzes the AV user opinion contained in the response RA and determines whether the AV user notification is positive, negative or undetermined.

If the server S can not determine whether the AV user notification is positive or negative, the AV user notification is of indeterminate type and the process continues with a new step E28, with a step E50 described later in the description or by a step E52 described in the following description. If the AV user notification is a negative opinion, the process continues with a new step E28, a step E50 described in the following description or by a step E52 described in the following description. If the AV user notification is a positive one, the S server determines one or more ORD relationship between the document for which a notice was issued in the RA response and the other targeted documents in the RP response; other documents selected in step E30. An ORD order relationship is a relationship between the document for which the AV user notification is formulated and another document targeted in the response RP to the first RQ request. A targeted document in the response RP is a document selected by the server S in the step E30, that is to say a document for which a set of metadata is contained in the response RP. For example, the server S determines a first order relation between D10 and D5 and a second order relation between D10 and D77. An ORD order relationship is defined by a first document, a second document, and an operator defining a relationship between the two documents. The operator is for example "relevance greater than". An ORD order relation between a document for which the user notification AU is issued, for example D10, and another document selected in the step E30, for example D77, is for example of the type: (relevance of the document for which AV user notification is issued) greater than (relevance of another document selected in step E30). For example: (Relevance of D10)> (relevance of D5) Alternatively, order relationships are also determined when the RA response is negative. As an alternative, the request for DA notice is of the type "which document corresponds best to your request? And is for example transmitted after sending the RP response (step 34), before or after the detection of one or more actions of the user to access one or more of the documents cited. In a step E46, ORM relationships between keywords contained in the first request RQ are determined from the ORD order relationships between documents. More specifically, a second value of relevance Vi is determined for each document Di contained in a document ORD relationship between documents, vis-à-vis the first request RQ. The value of relevance Vi of a document Di for the first request RQ is for example determined by the equation: Vi = Xk ln (1 + (e-1) nbr (Mk) nbd (Mk)) MkERQnDi in which Mk is a key word of the request RQ contained in the document Di, Xk the weighting value associated with the keyword Mk, ln is the natural logarithm, e is the constant of Neper, nbr (Mk) is the number of occurrences of the word Mk in the request RQ and nbd (Mk) is the number of occurrences of the word Mk in the description associated with the document Di. In the equation above, the weighting value Xk of a keyword Mk is an unknown. Then each ORD relation between two documents is transformed into a relation between two second weighting values. The relation of order "Di more relevant than Dj" is converted into a relation of the type "Vi> Vj". An ORM order is thus obtained between keywords of the type: Xk ln (1 + (e-1) nbr (Mk) nbd (Mk)) MkERQnDi> Xk ln (1 + (e-1) nbr (Mk) nbd (Mk)) MkERQnDi The determined ORM order relationships are stored in a memory accessible by the server S, for example a table of the relations TR. An order relationship between keywords is a defined inequality of keyword weighting values. As described later in this description, the ORM order relationships are then used to determine the weighting value of one or more keywords. The steps E28 to E46 are repeated for other user requests issued by the same terminal or by another terminal of the plurality of terminals. This gives a plurality of ORM order relationships between keywords. In a step E50, the server S determines a weighting value P1 for one or more keywords stored in the table of the keywords TAB according to ORM order relationships recorded in the table of relations TR. The weighting values are, for example, determined by a Support Vector Machine (SVM) applying a classification algorithm to the registered ORM order relations. Christopher J.C. Burges' document "A Tutorial on Support Vector Machines for Pattern Recognition" (Data Mining and Knowledge Discovery - 1998 - Volume 2 - pages 121 to 167) describes an example of SVM. The SVMs are then statistical tools used to determine the weighting values Xk of the keywords Mk.They make it possible to maximize the satisfaction of the order relationships between ORM documents The weighting values Pk of the keywords Mk are the Xk values thus determined.

The algorithm therefore makes it possible to assign a high weighting value to a keyword contained in a document whose relevance is considered high and instead to assign a low weighting value to a keyword contained in a document whose relevance is estimated low. Each weighting value Pk thus obtained for a keyword Mk is then recorded in association with the keyword Mk in the table of keywords TAB. As an alternative, in step E44, the ORD order relationships are recorded in the relationship table TR in association with the key words of the first request RQ and the step E46 for determining the ORM relations between words keys is performed for each ORD order relationship between documents just prior to performing step E50 of determining the weighting values. During a step E52 (FIG. 4b), a user request RQ2 containing one or more keywords is transmitted by a terminal of the plurality, for example the terminal T3, to the server S. The request RQ2 represents a current request.

The request RQ2 is received by the server S during a step E54. During a step E56, a relevance value VPi is determined, for the current request RQ2, for each document Di of the plurality of documents. The relevance value VPi represents the relevance of a document Di considered with respect to the current request RQ2.

The value of relevance VPi determined for a document Di recorded in a database BD is calculated on the one hand according to at least one weighting value Pj associated with at least one key word Mj extracted from the current request RQ2 and associated to the document Di and secondly according to at least one occurrence number of the at least one keyword in the description associated with the document Di.

The relevance value VPi of a document Di for the current request RQ2 is for example determined by the equation: VPi = Pj ln (1 + (e-1) nbr (Mpnbd (MD) MjERQ2nDi in which Mj is a key word, Pj the weighting value associated with the keyword Mj, ln is the natural logarithm, e is the Neper constant, nbr (Mj) is the number of occurrences of the word Mj in the current request RQ2 and nbd (Mj ) is the number of occurrences of the word Mj in the description associated with the document Di.

As an alternative, the value of relevance VPi determined for a document Di recorded in a database BD is calculated according to at least one weighting value associated with at least one keyword of the current request RQ2 associated with the document Di and does not take into account the number of occurrences. In a step E58, documents are selected from the plurality of documents based on the calculated values of relevance VPi. For example, the 3 documents Dx, Dy, Dz for which the relevance value is the highest are selected. As an alternative, only documents for which the relevance value is greater than a predefined threshold are selected.

More generally, the documents are selected according to one or more predefined criteria relating to the relevance value of the documents of the plurality of documents. Then, during a step E60, a response RP2 to the current request RQ2 is transmitted by the server S to the terminal T3.

The response RP2 contains for each document selected in step E58, all or part of all the metadata associated with the document. In the embodiment described, the response RP2 contains for each selected document, a URL address of the document. The response RP2 is received by the terminal T3 having transmitted the request RQ2 and its content, that is to say the URLs, is displayed on a screen of the terminal T3 during a step E62. In a step E64, a user of the terminal T3 selects a URL from the URLs displayed on the screen. For example, the selection is made in a known manner by clicking on the screen at the position on the screen where the URL is displayed by means of a computer mouse. This selection represents a user action AC2 vis-à-vis the associated document. In known manner, the "click" on the URL causes the terminal to send a request for access to the document associated with the selected URL, for example the document Dz. The document Dz is received by the terminal T3, in response to the "click" and displayed on the screen of the terminal T3. During a step E66, the server S detects the user action AC2 and during a step E68, the server S sends to the terminal T3, a request for user notification DA2 relating to the document Dz. The user request request DA2 is for example a text message asking the user if the document Dz is relevant to the current request RQ2. During a step E70, the terminal T3 sends the server S a response RA2 to the request for notification DA2. The RA2 response contains a user-defined AV2 notice. The response RA2 is received by the server S during a step E72. In a step E74, the server S analyzes the response received and determines, based on this response, one or more ORD relationship between the document for which a notice has been issued, here the document Dz, and the others targeted documents in response RP2, that is, the other documents selected in step E58. Step E74 is similar to step E44 described above. Step E74 is followed by a step E76 of determining ORM order relationships between keywords of request RQ2. Step E76 is similar to step E46 described above. The order relationships between determined ORM keywords are recorded in the relationship table TR. They are for example added to the ORM order relationships recorded during the execution of steps E46 carried out previously.

As an alternative, the ORM relations cooperating in the determination of weighting values are erased and the order relationships determined in step E76 contribute to the establishment of a new list of relations of order. As an alternative, the request for an opinion is of the type "which document corresponds best to your request? And is for example transmitted after the sending of the response RP2, before or after the detection of one or more clicks of the user. Steps E54 to E76 are reiterated by the server S for other common user requests issued by the same terminal or by another terminal of the plurality of terminals. Subsequently, during a step E78, the weighting values Pi of one or more Mi key words are updated from the order relationships between registered ORM keywords.

Registered ORM order relationships are used at the input of an SVM to determine new weight values. Step E78 is similar to step E50 previously described. Step E78 is for example performed periodically, for example once a day or once a week.

In the embodiments described, the database is indifferently a database containing predefined documents, that is to say a database whose content does not evolve or a database whose content is regularly updated. . Thus, at any point in the process, documents can be added to the database BD. Similarly, documents can be removed from this database. In the described embodiment, the order relationships between ORM keywords are maintained in the TR relationship table and used to update the keyword weight values. These relationships are therefore used even when the documents used to determine these relationships have been removed from the database BD. FIG. 5 represents a server S according to one embodiment of the invention. The server S comprises a reception module of a current request RRQ, a document selection module SEL and a transmission module TRP of a response to a current request. The reception module of a current request RRQ is able to receive a request for documents containing one or more keywords transmitted by a remote user terminal. The selection module SEL is able to select among the documents of the plurality, at least one document according to at least one value of relevance determined for the at least one document as a function of at least one weighting value associated with the document. at least one keyword of said current request associated with said document, said weighting value being predetermined as a function of at least one user notification relating to at least one response to at least a first request prior to said current request.

The transmission module TRP is able to transmit a response to the current request containing at least one set of metadata relating to at least one selected document. The server S may also comprise a determination module DET adapted to determine at least one weighting value of at least one keyword according to at least one user notification relating to at least one response to at least a first request prior to said current request. The server S embodying the invention is for example a microcomputer which comprises, in a known manner, in particular a processing unit MT equipped with a microprocessor, a read-only memory of the ROM type, a random access memory of the RAM type. The server S may comprise in a conventional and non-exhaustive manner the following elements: a keyboard, a screen, a microphone, a speaker, a communication interface, a disk drive, a storage medium ... The read-only memory of ROM type includes registers storing a computer program PG1 including program instructions adapted to implement a method of transmitting at least one set of metadata associated with at least one of a plurality of documents described with reference to the Figures 2, 4a and 4b. This program PG1 is thus adapted to receive a current request comprising at least one key word, to select from among the documents of the plurality of documents, at least one document based on at least one value of relevance determined for the said document. least one document based on at least one weighting value associated with at least one keyword of said current request associated with said document, said weighting value being predetermined according to at least one user notification relating to at least one response to at least a first request prior to said current request, and to transmit a response to said current request containing at least one set of metadata relating to at least one selected document.

The program PG1 is also adapted to determine at least one weighting value of at least one keyword based on at least one user notification relating to at least one response to at least a first request prior to said current request. When powering on, the program PG1 stored in the ROM is transferred into the RAM which will then contain the executable code of the invention and registers to store the variables necessary for the implementation of the invention. . More generally, a storage means, readable by a computer or by a microprocessor, integrated or not to the device, possibly removable, stores a program implementing the transmission method.

FIG. 6 represents a terminal T according to one embodiment of the invention. The terminal T comprises a transmission module of a current request TRQ and a reception module RRP of a response to a current request. The transmission module of a current request TRQ is able to transmit a request for documents containing at least one keyword to a server S.

The reception module RRP is able to receive a response to the current request containing at least one set of metadata relating to at least one document, said document being selected from the documents of the plurality of documents, according to at least one value of relevance determined for said at least one document based on at least one weighting value associated with at least one keyword of said current request associated with said document, said weighting value being predetermined according to at least one relative user opinion at least one response to at least a first request prior to said current request. The terminal T embodying the invention is for example a microcomputer which comprises, in a known manner, in particular a processing unit UT equipped with a microprocessor, a read-only memory of the ROM type, a random access memory of the RAM type, a ECR screen. The terminal T may comprise in a conventional and non-exhaustive manner the following elements: a keyboard, a microphone, a speaker, a communication interface, a disk drive, a storage means, etc. The read-only memory of ROM type includes registers storing a computer program PG2 having program instructions adapted to implement a method of obtaining at least one set of metadata associated with at least one document selected from a plurality of documents described in reference to Figures 2, 4a and 4b. This program PG2 is thus adapted to transmit a current request comprising at least one keyword and to receive a response to said current request containing at least one set of metadata relating to at least one document selected from the documents of the plurality of documents based on at least one determined relevance value for said at least one document as a function of at least one weighting value associated with at least one keyword of said current request associated with said document, said weighting value being predetermined as a function of at least one user notification relating to at least one response to at least a first request prior to said current request. More generally, a storage means, readable by a computer or by a microprocessor, integrated or not to the device, possibly removable, stores a program implementing the transmission method.

The invention is implemented by means of software and / or hardware components. In this context, the term "module" may correspond in this document to a software component, to a hardware component or to a set of hardware and / or software components, capable of implementing a function or set of functions, as described below for the module concerned.

A software component corresponds to one or more computer programs, one or more subroutines of a program, or more generally to any element of a program or software. Such software component is stored in memory then loaded and executed by a data processor of a physical entity (terminal, server, gateway, settop-box, router, etc.) and is likely to access the hardware resources of this physical entity (memories, recording media, communication bus, input / output electronic boards, user interfaces, etc.). In the same way, a material component corresponds to any element of a material set (or hardware). It may be a programmable hardware component or with an integrated processor for running software, for example an integrated circuit, a smart card, an electronic card for executing a firmware, etc.

Claims (7)

REVENDICATIONS1. Method for obtaining at least one set of metadata associated with at least one document selected from a plurality of documents, characterized in that it comprises the following steps: - transmission (E2, E52) of a current request (RQ1, RQ2) having at least one keyword; reception (E12, E62) of a response (RP1, RP2) to said current request containing at least one set of metadata associated with at least one document, said at least one document (Di) being selected from the documents of the plurality of documents according to at least one relevance value (VPi) determined for said at least one document (Di) as a function of at least one weighting value (Pj) associated with at least one keyword (Mj) of said current request (RQ1, RQ2) associated with said document (Di), said weighting value (VPi) being predetermined according to at least one user notification (DA) relating to at least one response (RP) to at least a first request (RQ) prior to said current request. 2. Method of obtaining according to claim 1 characterized in that it comprises a step of displaying (E62) metadata extracted from said at least one received set followed by a step of transmitting a relative user notification (DA2). at least one targeted document in said response. 3 Method of obtaining according to claim 1 wherein said metadata comprises a pointer for accessing a document and wherein the method comprises a step of displaying said pointer, a step of accessing said document via said pointer, a step receiving a request for user notification for said document and a step of transmitting a notice on the relevance of said document in response to said request for user notification. Method for transmitting at least one set of metadata associated with at least one document selected from among a plurality of documents, characterized in that it comprises the following steps: reception (E4, E54) of a current request (RQ1, RQ2) having at least one keyword; selecting (E8, E58) among the documents of the plurality of at least one document as a function of at least one determined value of relevance for said at least one document as a function of at least one weighting value associated with at least one document a key word of said current request associated with said document, said weighting value being predetermined as a function of at least one user notification relating to at least one response to at least a first request prior to said current request; - transmitting (E10, E60) a response (RP1, RP2) to said current request containing at least one set of metadata relating to at least one selected document. The transmission method of claim 4 wherein said relevance value is further determined based on at least one occurrence number of said at least one query keyword for said at least one document. 6 transmission method according to claim 4 characterized in that it comprises the following steps, performed prior to said step of receiving a current request: - receiving steps (E28) from a plurality of terminals, first queries (RQs) containing at least one keyword; steps of transmission (E32) of responses (RP) to the first requests, respectively comprising at least one set of metadata associated with at least one document selected from the plurality of documents as a function of at least one keyword of a first query; - User notification (AV) reception steps (E42) for at least one of the transmitted responses; a step of determining (E50) at least one weighting value of at least one keyword according to at least one user notification (AV) received. 7 transmission method according to claim 6 wherein the user opinion relates to the relevance of a targeted document in a response to a first request and wherein the step of receiving the notice is followed by a step of determining at least one order relationship (ORD) between said target document and another targeted document in said response and wherein said at least one weighting value of at least one keyword is determined based on at least one 8 A transmission method according to claim 4, characterized in that the step of transmitting a response to a current request is followed by the following steps: detection (E66) of a user action (AC2) a document targeted in the response (RP2); - transmission (E68) of a request for user notification (DA2) following said detection of a user action; - receiving (E72) a user notification (AV2) relating to said document; -22- - determining (E76) at least one order relation (ORM) between at least one keyword of the current request associated with said document and at least one keyword of the current request associated with another targeted document in said response; - Recording of said at least one specific order relationship 9 Transmission method according to claim 8 characterized in that it comprises a step (E78) for updating at least one weighting value of at least one word key according to at least one registered order relationship (ORM). 10. Terminal (T, T3) characterized in that it comprises: - a transmission module (TRQ) of a current request (RQ1, RQ2) comprising at least one keyword; a reception module (RRP) for a response (RP1, RP2) to said current request containing at least one set of metadata associated with at least one document, said at least one document being selected from documents of a plurality of documents according to at least one value of relevance determined for said at least one document as a function of at least one weighting value associated with at least one keyword of said current request associated with said document, said weighting value being predetermined by function of at least one user notification relating to at least one response to at least a first request prior to said current request. 11. Server (S) characterized in that it comprises: - a reception module (RRQ) of a current request (RP1, RP2) comprising at least one keyword; a selection module (SEL) among documents of a plurality of documents of at least one document as a function of at least one value of relevance determined for the said at least one document as a function of at least one weighting value associated with at least one keyword of said current request associated with said document, said weighting value being predetermined according to at least one user notification or at least one user action relating to at least one response to at least a first request prior to said current request; a module for transmitting (TRP) a response (RP1, RP2) to said current request containing at least one set of metadata relating to at least one selected document.