Techniques for knowledge discovery by constructing knowledge correlations using concepts or terms

• US 8,126,890 B2
• Filed: 12/21/2005
• Issued: 02/28/2012
• Est. Priority Date: 12/21/2004
• Status: Active Grant

• Current Assignee: Make Sence Incorporated
• Original Assignee: Make Sence Incorporated
• Inventors: Bobick, Mark, Wimmer, Carl
• Primary Examiner(s): Yen, Syling
• Application Number:
  US11/314,835
• Publication Number:
  US 20060167931A1
• Time in Patent Office: 2,260 Days
• Field of Search: 707/3, 707101-102, 707/706, 707/722, 707/736, 707/737, 707/741, 707/746, 707/748, 707/758, 707/781, 707999003- 7, 709/202, 706/46, 704/9, 715/256
• US Class Current: 707/736
• CPC Class Codes: G06F 16/36 Creation of semantic tools, e.g. ontology or thesauri
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06N 5/022 Knowledge engineering; Knowledge acquisition

Abstract

Techniques for identifying knowledge use a graphical user interface for inputting one or more terms to be explored for additional knowledge. Then a search is conducted across one or more sources of information to identify resources containing information about or information associated with said terms. The resources are decomposed into elemental units of information and stored in a data structures called nodes. A group of nodes are stored in a node pool and, from the node pool, correlations of nodes are constructed that represent knowledge using information about relation types. Information about relations types is determined using a relation classifier.

Claims

1. A method for identifying knowledge comprising:

using an input device coupled to a processor and associated memory for permitting inputting at least one term to be explored for additional knowledge;

using the processor and the associated memory coupled to the input device to identify resources comprising information associated with the at least one term;

using the processor and the associated memory for decomposing the resources into a plurality of nodes by at least

receiving each resource as a sequential plurality of sentences,

processing each sentence to determine whether a given sentence matches a syntactical pattern, and

if the given sentence matches the syntactical pattern, then processing the given sentence into at least one node including a subject, an attribute, and a word based relational bond therebetween;

using the processor and the memory for storing the plurality of nodes in a node pool; and

using the processor and the associated memory for constructing and storing correlations of linked together nodes of the node pool representing knowledge associated with the at least one term to be explored by at least iteratively adding nodes from the node pool onto an end of a chain of nodes by searching the node pool for a match between an attribute of a chained node and a subject of another unchained node in the node pool.

2. The method of claim 1 wherein the decomposing includes mapping the syntactical pattern to the subject, the attribute, and the word based relational bond of each node in the node pool.

3. The method of claim 2 wherein at least one of a verb and an adverb of the syntactical pattern is mapped to the word based relational bond of each node in the node pool.

4. The method of claim 1 further comprising using the processor and the memory for presenting to users the knowledge in the form of completed correlations.

5. The method of claim 1 wherein the decomposing includes constructing the correlations of linked together nodes using an association function and a relation classifier for determining whether a given sentence should be formed into a node.

6. The method of claim 3 wherein the relation classifier provides information about a relation case.

7. The method of claim 5 wherein the relation classifier uses a lexicon of verbs and adverbs.

8. The method of claim 5 wherein the relation classifier uses a map of verbs and adverbs to specific cases of relation.

9. The method of claim 7 wherein the relation classifier uses the verbs and adverbs contained in a lexicon to map to specific cases of relation.

10. The method of claim 9 wherein an independent map of verbs and adverbs to specific cases of relation is supplied for each language of the plurality of resources.

11. The method of claim 6 wherein the relation case of each node is determined by examination of the word based relational bond of a respective node.

12. The method of claim 5 wherein the association function uses relation case information to pre-select from the node pool all information nodes eligible for association.

13. The method of claim 1 wherein the word based relational bond of each node correlates to a relation comprising at least one of:

an extensional relation;

an intentional relation;

a class relation;

a mereological relation;

a topological relation;

an existential relation;

an action relation;

a transitional relation;

a causal relation;

a dependency relation;

a semiotic relation;

a mediated relation;

a conventional relation; and

a property-based relation.

14. The method of claim 13 wherein the decomposing includes performing a disambiguation function to resolve an ambiguous case for a given node when the word based relational bond thereof correlates to a plurality of relations.

15. The method of claim 14 wherein the disambiguation function examines the word based relational bond of the given node to resolve the ambiguous case of relation.

16. The method of claim 14 wherein the disambiguation function changes the word based relational bond of the given node to resolve the ambiguous case of relation.

17. The method of claim 14 wherein the disambiguation function creates at least one additional node representing at least one relation of the plurality of relations for the word based relational bond of the given node to resolve the ambiguous case of relation.

18. The method of claim 5 wherein the relation classifier uses a relation of a proposed node to determine if the proposed node should be constructed.

19. The method of claim 18 wherein the relation classifier performs a lookup in a lexicon of verbs and adverbs using as an argument the relation of the proposed node to determine if the proposed node should be constructed.

20. The method of claim 19 wherein the relation classifier permits the proposed node to be constructed if the lookup is successful, and the relation classifier blocks the construction of the proposed node if the lookup fails.

21. The method of claim 5 wherein the association function processes a plurality of parts of an origin node to determine if the origin node can be associated with a candidate node, the plurality of parts comprising at least two of a first relatum, a relation, a second relatum, and a source.

22. The method of claim 21 wherein the association function uses additional tests to determine if the origin node can be associated with the candidate node.

23. The method of claim 22 wherein the additional tests include finding a match between a relatum of the candidate node and a second relatum of the origin node or terminus node.

24. The method of claim 22 wherein the additional tests include finding a version of a relatum of the candidate node and a second relatum of the origin node or terminus node.

25. The method of claim 22 wherein the additional tests further include finding a match between a synonym of a relatum of the candidate node and a second relatum of the origin node or terminus node.

26. The method of claim 22 wherein the additional tests further include finding a match between terms from a thesaurus that are related to a relatum of the candidate node and a second relatum of the origin node or terminus node.

27. The method of claim 22 wherein the additional tests further include finding a match between terms in a dictionary definition of a relatum of the candidate node and a second relatum of the origin node or terminus node.

28. The method of claim 22 wherein the additional tests further include finding a match between terms in an authoritative source between a relatum of the candidate node and a second relatum of the origin node or terminus node.

29. The method of claim 22 wherein the additional tests further include finding a match between terms in an entry of an encyclopedia source including a relatum of the candidate node and a second relatum of the origin node or terminus node.

30. The method of claim 22 wherein the additional tests further include finding terms from at least one of a taxonomy, an ontology, and a semantic net adjacent to a relatum of the candidate node and that match a second relatum of the origin node or terminus node.

31. The method of claim 22 wherein the additional tests further include finding terms in a relatum of the candidate node related to a second relatum of the origin node or terminus node where the relation between content of the relatum of the candidate node and the content of the second relatum of the origin node or terminus node is established by an authoritative source.

32. The method of claim 22 wherein the additional tests further include finding a match between terms in a definition in an authoritative source of a relatum of the candidate node and a second relatum of the origin node or terminus node.

33. A physical computer-readable medium having instructions stored thereon which, when executed by a computer, cause the computer to perform an method for identifying knowledge comprising:

permitting inputting at least one term to be explored for additional knowledge;

identifying resources comprising information associated with the at least one term;

decomposing the resources into a plurality of nodes by at least

receiving each resource as a sequential plurality of sentences,

processing each sentence to determine whether a given sentence matches a syntactical pattern, and

if the given sentence matches the syntactical pattern, then processing the given sentence into at least one node including a subject, an attribute, and a word based relational bond therebetween;

storing the plurality of nodes in a node pool; and

constructing and storing correlations of linked together nodes of the node pool representing knowledge associated with the at least one term to be explored by at least iteratively adding nodes from the node pool onto an end of a chain of nodes by searching the node pool for a match between an attribute of a chained node and a subject of another unchained node in the node pool.

34. The physical computer-readable medium of claim 33 wherein the decomposing includes mapping the syntactical pattern to the subject, the attribute, and the word based relational bond of each node in the node pool.

35. The physical computer-readable medium of claim 34 wherein at least one of a verb and an adverb of the syntactical pattern is mapped to the word based relational bond of each node in the node pool.

36. The physical computer-readable medium of claim 33 wherein the decomposing includes constructing the correlations of linked together nodes using an association function and a relation classifier for determining whether a given sentence should be formed into a node.

37. The physical computer-readable medium of claim 35 wherein the relation classifier provides information about a relation case.

38. The physical computer-readable medium of claim 36 wherein the relation classifier uses a lexicon of verbs and adverbs.

39. The physical computer-readable medium of claim 36 wherein the relation classifier uses a map of verbs and adverbs to specific cases of relation.

40. The physical computer-readable medium of claim 38 wherein the relation classifier uses the verbs and adverbs contained in a lexicon to map to specific cases of relation.

41. An apparatus for identifying knowledge comprising:

an input device for permitting inputting at least one term to be explored for additional knowledge; and

a processor and memory associated therewith and coupled to said input device for

identifying resources comprising information associated with the at least one term,

decomposing the resources into a plurality of nodes by at least

receiving each resource as a sequential plurality of sentences,

processing each sentence to determine whether a given sentence matches a syntactical pattern, and

if the given sentence matches the syntactical pattern, then processing the given sentence into at least one node including a subject, an attribute, and a word based relational bond therebetween;

storing the plurality of nodes in a node pool, and

constructing and storing correlations of linked together nodes of the node pool representing knowledge associated with the at least one term to be explored by at least iteratively adding nodes from the node pool onto an end of a chain of nodes by searching the node pool for a match between an attribute of a chained node and a subject of another unchained node in the node pool.

42. The apparatus of claim 41 wherein the decomposing includes mapping the syntactical pattern to the subject, the attribute, and the word based relational bond of each node in the node pool.

43. The apparatus of claim 42 wherein at least one of a verb and an adverb of the syntactical pattern is mapped to the word based relational bond of each node in the node pool.

44. The apparatus of claim 41 wherein the decomposing includes constructing the correlations of linked together nodes using an association function and a relation classifier for determining whether a given sentence should be formed into a node.