Self-organization approach to semantic interoperability in peer-to-peer information exchange

• US 20060074906A1
• Filed: 10/04/2005
• Published: 04/06/2006
• Est. Priority Date: 10/05/2004
• Status: Active Grant

• Current Assignee: Sony Europe (Amsterdam) BV (Sony Group Corp.)
• Original Assignee: Sony France SA (Sony Group Corp.)
• Inventors: Steels, Luc, Hanape, Peter
• Granted Patent: US 7,707,147 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/907 Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
• CPC Class Codes: G06F 16/908 using metadata automatically derived from the content
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques

Abstract

In the exchange of information in a peer-to-peer information system network, the problem of semantic interoperability of the communication system used by the agent associated with every peer is addressed not by imposing a universal pre-defined ontology over universally defined conceptual schemata, but instead using mechanisms, inspired from natural language, that enable each agent to develop a repertoire of grounded categories and labels for these categories and to negotiate their use and semantics with other agents. The communication system, as well as its semantics, is hence emergent and adaptive instead of predefined.

Claims

1. An information system comprising:

means for accessing a first collection of data items;

retrieval-request means adapted to enable a user of said information system to identify a class of data items said user desires to retrieve from a remote information system, wherein the user identifies said class of data items by reference to a set of one or more examples, said examples being data items in said first collection;

an information agent responsive to the retrieval-request means and adapted to prepare an initial query for transmitting the user'"'"'s request to a remote information system, said query including a label designating a data item-category desired for retrieval, said information agent comprising a classifier adapted automatically to assign a category to said set of examples identified by the user, the information agent being adapted to include in the initial query a label corresponding to the assigned category;

output means for outputting the initial query prepared by the information agent;

input means for receiving information identifying data items retrieved by a remote information system in response to a query output by said output means;

presentation means for presenting the first user with said information identifying the data items retrieved by the remote information system; and

selection means for enabling the first user to make a selection from among the data items retrieved from the remote information system in response to a query;

characterized in that;

the information agent maintains an inventory of categories, labels, and bindings associating labels and categories with each other, and is adapted to include, in the initial query, the label having the strongest binding to the assigned category; and

the information agent is responsive to the selection means whereby to change the strength of bindings between labels and categories dependent upon the user'"'"'s selection.

2. The information system of claim 1, wherein the information agent is adapted:

to determine whether the quantity of data items selected by the user from among the retrieved data items is greater or lower than a predetermined threshold; and

if the selected quantity is greater than the predetermined threshold, to increase the strength of the binding between the label used in the initial query and the assigned category, relative to the strength of the bindings between other labels and said assigned category.

3. The information system of claim 1, wherein the information agent is adapted:

to determine whether the quantity of data items selected by the user from among the retrieved data items is greater or lower than a predetermined threshold; and

if the selected quantity is lower than the predetermined threshold, to decrease the strength of the binding between the label used in the initial query and the assigned category.

4. The information system of claim 1, comprising:

structuring means enabling the user to define a data structure applicable to data items of said first collection, the nodes of said data structure being labeled according to a user taxonomy;

wherein the retrieval-request means is adapted to enable a user of said information system to identify, by reference to the user taxonomy, the class of data item desired for retrieval.

5. The information system of claim 1, wherein each of the categories in the list maintained by the information agent is defined in terms of a distinctive feature set and the classifier is adapted to determine the category to which the user-identified set of data items should be assigned by comparing the features of the user-identified set of examples with the distinctive feature sets of the categories in the list maintained by the information agent.

6. The information system of claim 1, and comprising feedback means responsive to the selection means whereby to cause a signal identifying the user'"'"'s selection to be output to the remote information system.

7. The information system of claim 1, wherein the information agent is adapted to cause the output, to the remote information system, of data identifying the user-identified set of examples, in the case where the information received from the remote information system indicates that no data items have been retrieved by said remote information system in response to a query, or in the case where the user operates the selection means to select a quantity of retrieved data items lower than a predetermined threshold.

8. The information system of any one of claim 1, wherein the information agent is adapted:

to determine whether the quantity of data items selected by the user from among the retrieved data items is greater or lower than a predetermined threshold;

if the selected quantity is below the threshold, to cause the classifier to determine a revised category applicable to the aggregate of the examples identified to the retrieval-request means and the data items selected by the user from among the data items retrieved by the remote information system; and

if the revised category is different from the assigned category that was labeled in the initial query, to prepare and output to the remote information system a complement to the initial query, said complement including a revised label being the label having the strongest binding to the revised category.

15. A peer-to-peer information-exchange network comprising a plurality of peer information systems operable in a client mode or a server mode, each information system operating in client mode being an information system according to claim 1, and each information system operating in server mode being an information system.

9. An information system comprising:

means for accessing a second collection of data items;

input means for receiving an initial query from a remote information system, said query requesting retrieval of a data items from said second collection and including a label;

an information agent comprising decoding means for processing the received initial query whereby to determine, based on said label, an initial category of data item for retrieval from said second collection, said information agent being adapted to prepare an initial response identifying the data items in said second collection belonging to said determined initial category;

output means for outputting said initial response to the remote information system;

characterized in that;

the information agent is adapted to maintain an inventory of categories, labels, and bindings associating labels and categories with each other, and the decoding means is adapted to determine the initial category of data items for retrieval, by selecting the category having the strongest binding to the label received in the initial query, and

the information agent is responsive to feedback received from the remote information system whereby to change the strength of bindings between labels and categories, said feedback being indicative of a selection, made by a user of said remote information system, from among the data items identified in said initial response.

10. The information system of claim 9, wherein the information agent is adapted:

to determine whether the feedback received from the remote information system indicates that quantity of data items selected by the user of the remote information system is greater or lower than a predetermined threshold; and

if the selected quantity is greater than the predetermined threshold, to increase the strength of the binding between the label in the received initial query and the initial category determined by the decoder, relative to the strength of the bindings between said label and other categories.

11. The information system of claim 9, wherein the information agent is adapted:

to determine whether the feedback received from the remote information system indicates that quantity of data items selected by the user of the remote information system is greater or lower than a predetermined threshold; and

if the selected quantity is lower than the predetermined threshold, to decrease the strength of the binding between the label in the received initial query and the initial category determined by the decoder.

12. The information system of claim 9, wherein each of the categories in the list maintained by the information agent is defined in terms of a distinctive feature set and the information agent is adapted to identify data items in said second collection, for retrieval, by comparing the features of said data items with the distinctive feature set of the determined initial category.

13. The information system of any one of claim 9, wherein:

each of the categories in the list maintained by the information agent is defined in terms of a distinctive feature set; and

the information agent comprises a classifier adapted to identify a superceding category of data item for retrieval, by analysis of example data received from said remote information system, said example data identifying a set of data items constituting examples of the class of data items desired for retrieval, said classifier being adapted to identify said superceding category for retrieval by comparing the features of the example data items with the distinctive feature sets of the categories in the list maintained by the information agent.

14. The information system of claim 13, wherein the classifier is adapted such that, if none of the feature sets of the categories in the list maintained by the information agent is distinctive with regard to the example data items, the classifier determines the distinctive feature set of said example data items and causes a corresponding new category to be created.

16. An information system comprising:

means for accessing a first collection of data items;

specifying means adapted to enable a user of said information system to identify a class of data items said user desires to identify to a remote information system, wherein the user identifies said class of data items by reference to a set of one or more examples, said examples being data items in said first collection;

an information agent responsive to the specifying means and adapted to prepare an initial message for transmission to a remote information system, said initial message including a label designating a data-item-category, said information agent comprising a classifier adapted automatically to assign a category to said set of examples identified by the user, the information agent being adapted to include in the initial message a label corresponding to the assigned category;

output means for outputting the initial message prepared by the information agent; and

input means for receiving from a remote information system a request for identification of data items, in said first data collection, in the category indicated by the label included in the label of the initial message;

wherein the information agent is adapted to identify said set of examples to said remote information system;

characterized in that;

the information agent is adapted to maintain an inventory of categories, labels, and bindings associating labels and categories with each other, and to include in the initial message the label having the strongest binding to the assigned category; and

the information agent is responsive to feedback received from the remote information system whereby to change the strength of bindings between labels and categories, said feedback being indicative of a selection, made by a user of said remote information system, from among said set of data items.

17. A method of managing information exchange in a peer-to-peer network comprising a plurality of peer information systems operable in a client mode or a server mode, each peer information system comprising:

a collection of data items, means for a user to indicate a class of data items desired for retrieval from a remote peer, and means for the user to make a selection from among data items retrieved by a remote peer, the method comprising the steps of;

providing each peer information system with an information agent maintaining an inventory of labels, categories, and bindings associating the labels and categories with each other,

wherein, when formulating a query in client mode, the information agent analyses example data representing the user'"'"'s desired class of data item whereby to determine a client-side category for the query, and includes in the query output to the remote peer the label having the strongest binding to said determined client-side category;

wherein, when responding to a query in server mode, the information agent decodes the label received in the query and determines, as the category of data item for retrieval, the server-side category having the strongest binding to the received label;

wherein the information agent is adapted to change the strength of the bindings between the labels and categories of its inventory dependent upon users'"'"' selections among data items retrieved from remote peers.