Object-based information storage, search and mining system method

• US 20060136402A1
• Filed: 12/20/2005
• Published: 06/22/2006
• Est. Priority Date: 12/22/2004
• Status: Abandoned Application

• Current Assignee: Tsu-Chang Lee
• Original Assignee: Tsu-Chang Lee
• Inventors: Lee, Tsu-Chang
• Application Number:
  US11/314,031
• Publication Number:
  US 20060136402A1
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/2465 Query processing support for facilitating data mining operations in structured databases
• CPC Class Codes: G06F 16/41 Indexing; Data structures therefor; Storage structures
• CPC Class Codes: G06F 16/5838 using colour
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation

Abstract

The patent application taught an object based information storage, search and mining system and method, which can store, retrieve, and mine massively amount of object data efficiently. A first embodiment includes an object based search engine which can organize and identify the objects efficiently from the search space based on certain operational measure quantifying the search objectives. A second embodiment includes an operational measure estimator which can adaptively adjust or reconfigure itself to improve the accuracy of the estimation. A third embodiment includes a method for mapping objects onto a topological maps to enable identifying objects rapidly from the search space according to certain measures for search objectives. A forth embodiment includes a method for conducting object based search through mapping query objects and search objects onto topological maps and searching effectively through the maps based on certain operational measures.

Claims

1. A method for conducting an object-based search comprising the steps of

determining from the query, a set of characteristics corresponding to the search objective for the query,

exploring objects in a search space, and

identifying objects optimally conforming to the search objective, wherein

conforming identified objects have a predetermined correspondence with the desired set of characteristics.

2. The method of claim 1, wherein the step of identifying includes, for each identified object,

extracting from the object a plurality of parameters,

determining from the parameters, the set of characteristics corresponding to the search objective, and

deriving a measure of conformability to the search objective based on the set of characteristics.

3. The method of claim 2 and further comprising estimating the measure of conformability of an expected outcome of the search.

4. The method of claim 3, and further comprising continually revising the estimated measure in the course of the search.

5. The method of claim 2, wherein the measure quantifies information associated with the search objective.

6. The method of claim 2, wherein the measure is an operational measure associated with a process for quantifying the search objective.

7. The method of claim 1, wherein objects in the search space are associated with nodes on a topological map (the Search Object Map—

SOM), and wherein the nodes are arranged to maximize correspondence between groups of objects associated with proximately located nodes.

8. The method of claim 7, wherein the map topology is graphical.

9. The method of claim 7, wherein the map topology is an N-dimensional lattice.

10. The method of claim 7, wherein the map is hierarchical.

11. The method of claim 7 wherein the step of identifying includes

locating at least a neighborhood on the topological map, each neighborhood comprising a group of correlated objects, wherein the neighborhood includes objects meeting a threshold correspondence with the set of characteristics, and

searching the neighborhood for objects optimizing the search objective.

12. The method of claim 11 wherein the step of identifying further includes searching beyond the neighborhood when no object in any of the neighborhood has the predetermined correspondence.

13. The method of claim 7, wherein the query is an object.

14. The method of claim 13, wherein objects in the history of queries are associated with nodes on a topological map (the Query Object Map), wherein the nodes are arranged to maximize correspondence between groups of query objects associated with proximately located nodes.

15. The method of claim 14, wherein the results of the past queries are recorded on the QOM as node locations on the SOM (the Search Object Map), corresponding to the resulting objects for the past queries.

16. The method of claim 15, wherein the step of identifying includes

locating the first neighborhood of the query object on the first topological object map (the Query Object Map), the neighborhood comprising a group of correlated past history query objects, wherein the neighborhood includes query objects meeting a threshold correspondence with the set of characteristics, and

for each past history query object in the first neighborhood, locating the node corresponding to its query result object on the second topological object map (the Search Object Map), and

searching the neighborhood for the nodes corresponding to past query result objects on the second topological object map (the Search Object Map) optimizing the search objective.

17. A multimedia object miner, comprising

a query object mapper for receiving and organizing query objects from service requesting agent, and

a search object mapper for retrieving and organizing search objects from one or more search spaces, and

an operational measure estimator for receiving a priori parameters and generating estimation for operation measure of an expected search objective, and

an object mining processor for controlling searches of the one or more search spaces and adaptively modifying parameters controlling operation of the query object mapper and the search object mapper.

18. The multimedia object miner of claim 17, wherein the query object mapper includes

a multimedia object buffer for maintaining query objects, and

an query object feature vector extractor for generating feature vectors from query objects, and

one or more topological maps for mapping a plurality of query objects based on the feature vectors.

19. The multimedia object miner of claim 17, wherein the search object mapper includes

a multimedia object buffer for maintaining search objects, and

an search object feature vector extractor for generating feature vectors from search objects, and

one or more topological maps for mapping a plurality of search objects based on the feature vectors.

20. The multimedia object miner of claim 17, wherein the operational measure estimator includes

a first buffer for storing the a priori data, and

a first parameter extractor for extracting the a priori search parameters from the a priori data,

a second buffer for storing the a posteri data,

a second parameter extractor for extracting the a posteri parameters from the a posteri data, and

an adaptive function estimator for providing an estimation of the operational measure of the search objective based on the a priori parameters,

wherein

the a posteri parameters are obtained from results of one or more preceding searches, and

the operation of the function estimator is reconfigurable based on the characteristics of the a posteri parameters and estimation error.

21. A method for mapping a multidimensional object on a topological map, comprising the steps of

locating a closest other object mapped on the topological map, the closest other object having a predetermined correspondence with the first object, and

mapping the object on the topological map in relative proximity to the other object, wherein

correspondence between objects is related to a difference measure associated with the objects.

22. The method of claim 21, the predetermined correspondence is a threshold difference measure, and wherein the step of locating includes

comparing a plurality of target objects with the object to obtain operational difference measures representing differences between the object and each of the plurality of target objects, and

selecting as the closest other object, a remaining target object with a lowest difference measure.

23. The method of claim 22, wherein the difference measure is an operational differential entropy, which measures the difference in information associated with the objects.

24. The method of claim 21 further comprising the step of

repeating the step of locating a selected number of times to identify one or more next closest objects, and

mapping the object on the topological map in relation to proximities of the object to the closest and the next closest objects.

25. The method of claim 21, wherein the topological map is a multidimensional lattice, and wherein the step of locating includes

extracting multidimensional feature vectors from objects, and

mapping the objects onto a lower-dimensional lattice according to difference measures of the multidimensional feature vectors.

26. The method of claim 25, wherein the step of mapping includes preserving neighborhood relationships between objects.

27. The method of claim 25, wherein at least one of the dimensions of the map is time-related.

28. The method of claim 25, wherein the n-dimensional lattice is adjusted through a SPAN algorithm.

29. The method of claim 21, the map is hierarchical, which further comprising the steps of obtaining one or more feature vectors of the input object,

locating a subset on the hierarchical map, the subset including objects having a predetermined correspondence with the input object, and

mapping the object within the subset,

wherein the hierarchical map includes a plurality of nodes.

30. The method of claim 29, wherein correspondence between objects is measured as an operational difference measure between the objects.

31. The method of claim 30, wherein the predetermined correspondence is a threshold difference measure, and wherein the step of locating includes

comparing a plurality of target objects to obtain relationships between the input object and each of the plurality of target objects, and

assigning to the subset, objects with the predetermined correspondence.

32. The method of claim 30, wherein the operational difference is measured as an operational differential entropy, wherein operational differential entropy quantifies difference in information contents between objects.

33. The method of claim 29, wherein the nodes include topological maps.