Computerized method and system for searching for text passages in text documents

• US 8,155,946 B2
• Filed: 12/22/2003
• Issued: 04/10/2012
• Est. Priority Date: 12/23/2002
• Status: Active Grant

• Current Assignee: Definiens AG (Astrazeneca PLC)
• Original Assignee: Definiens AG (Astrazeneca PLC)
• Inventors: Hudson, Richard, Jones, Dean, Klenk, Juergen, Schmidt, Guenter, Woischnik, Markus
• Primary Examiner(s): Vo, Huyen X.
• Application Number:
  US10/744,342
• Publication Number:
  US 20050278164A1
• Time in Patent Office: 3,032 Days
• Field of Search: 704 1- 10, 704/257, 704/270, 704/270.1, 704251-255, 707/3, 707/5, 706/55
• US Class Current: 704/4
• CPC Class Codes: G06F 16/3344 using natural language analysis
• CPC Class Codes: G06F 40/30 Semantic analysis

Abstract

A method is disclosed for searching for text passages in text documents. The method uses computerized prediction-based and ontology-based semantic language processing. Both the text passages that are to be located, as well as the text of the documents that are searched, are transformed on the basis of surface-syntactic and deep-syntactic information to generate a semantic network structure. The semantic network structure is linguistically processed based on information about other parts of the semantic network structure in order to improve the accuracy of the semantic network structure. Nodes in the semantic network structure are classified by linking the nodes in the semantic network structure to nodes in a pre-existing ontological network structure representing concepts in a language.

Claims

1. A computerized method for searching for text passages in text documents, comprising:

(a) parsing said text documents to obtain syntactic text information about said text documents;

(b) processing said syntactic text information to obtain semantic text information about said text documents, wherein said processing is performed by a computer;

(c) building said semantic text information into first semantic text networks;

(d) linking said first semantic text networks to an ontology;

wherein the ontology includes concepts and relationship slots, wherein each of the concepts defines an entity, and wherein each of the relationship slots defines a binary relation between two concepts;

(e) generating second semantic text networks from said text passages; and

(f) matching said first semantic text networks to second semantic text networks by specifying meanings of said text passages to find said text passages in said text documents.

2. The method of claim 1, wherein said syntactic text information includes surface-syntactic information and deep-syntactic information.

3. The method of claim 1, further comprising:

(g) linking said second semantic text networks to said ontology.

4. The method of claim 1, wherein said first semantic text networks and said second semantic text networks have the same meta structure.

5. The method of claim 1, wherein knowledge is extracted from said text documents as a result of said matching in (f).

6. The method of claim 5, wherein said knowledge is stored and can be searched.

7. The method of claim 5, wherein said knowledge is derived from nodes of said first semantic text networks according to rules specified in association with said second semantic text networks.

8. The method of claim 1, wherein said text documents include words, further comprising:

(g) classifying said words into lexemes and grammatical text elements.

9. The method of claim 8, further comprising:

(h) determining where said grammatical text elements are pointing to.

10. The method of claim 9, further comprising:

(i) building a chain of one or more of said grammatical text elements, said chain including a last grammatical text element, wherein each of said grammatical text elements in said chain, except said last grammatical text element, points to an additional grammatical text element, and wherein said last grammatical text element points to a lexeme.

11. The method of claim 8, wherein a chain comprises one or more grammatical text elements, further comprising:

(h) building a composition comprising a lexeme, one or more grammatical text elements, and one or more chains.

12. The method of claim 8, further comprising:

(h) classifying said lexemes into predicates and arguments.

13. The method of claim 12, wherein a predicate has a corresponding argument, and wherein said corresponding argument is classified to reflect a semantic role of said corresponding argument.

14. The method of claim 12, further comprising:

(i) classifying a relationship between a first lexeme and a second lexeme, wherein said relationship is a lexical government if said first lexeme is a predicate and said second lexeme is an argument, and wherein said relationship is a non-lexical government in all other cases.

15. The method of claim 8, further comprising:

(h) classifying said grammatical text elements into operators and coordinators.

16. The method of claim 8, wherein said words form associated nodes of said first semantic text networks, and wherein links between said associated nodes capture grammatical relationships between said words of said associated nodes, further comprising:

(h) classifying said grammatical text elements into operators and coordinators, wherein said coordinators link said associated nodes to form coordinated associated nodes, and wherein said coordinators form additional nodes that link an associated node to corresponding coordinated associated nodes.

17. The method of claim 1, wherein said text documents include words, wherein said words form associated nodes of said first semantic text networks, and wherein links between said associated nodes capture grammatical relationships between said words of said associated nodes.

18. The method of claim 1, wherein a first single word of said first semantic text networks is matched to a second single word of said second semantic text networks if said first single word and said second single word are identical or if said first single word is a synonym or a hyponym of said second single word.

19. The method of claim 1, wherein said first semantic text networks include a first text passage, wherein said first text passage includes words, wherein said second semantic text networks include words, wherein said first text passage is matched to said second semantic text networks if (i) said first text passage has a semantic structure identical to that of said second semantic text networks, (ii) corresponding words of said words of said first text passage and of said words of said second semantic text networks are classified identically, (iii) corresponding words of said words of said first text passage and of said words of said second semantic text networks are identical, are synonyms, or are hyponyms, or (iv) each link between associated nodes of said first semantic text networks is classified identically to a corresponding link between associated nodes of said second semantic text networks.

20. The method of claim 19, wherein said first text passage is determined using a template.

21. The method of claim 20, wherein said template is specified by a user.

22. The method of claim 21, wherein one of said words of said first text passage is selected for matching.

23. The method of claim 22, wherein said one of said words of said first text passage is matched based on a relationship slot of said ontology.

24. The method of claim 23, wherein said relationship slot is selected independently from said words of said first text passage.

25. The method of claim 24, wherein matching of words of said second semantic text networks produces results, and wherein said results are used in matching words of a group of additional semantic text networks.

26. The method of claim 22, wherein said first text passage has a context within said text documents, and wherein said context is used in said matching.

27. The method of claim 26, wherein a negation status and a modality status of said first text passage is determined.

28. The method of claim 27, wherein said first text passage includes multiwords, and wherein said multiwords are matched.

29. The method of claim 28, wherein a word of said second semantic text networks is a trigger word for starting said matching in (f).

30. A tangible computer-readable medium comprising computer-executable instructions for searching for text passages in text documents by performing the steps of:

(a) parsing said text documents to obtain syntactic text information about said text documents;

(b) processing said syntactic text information to obtain semantic text information about said text documents;

(c) building said semantic text information into first semantic text networks;

(d) linking said first semantic text networks to an ontology;

wherein the ontology includes concepts and relationship slots, wherein each of the concepts defines an entity, and wherein each of the relationship slots defines a binary relation between two concepts;

(e) generating second semantic text networks from said text passages; and

(f) matching said first semantic text networks to second semantic text networks by specifying meanings of said text passages to find said text passages in said text documents.

31. The computer-readable medium of claim 30, further for performing the step of:

(g) extracting knowledge from said text documents as a result of said matching in (f).

32. A computerized system for searching for text passages in text documents, comprising

means for parsing said text documents to obtain syntactic text information about said text documents;

means for processing said syntactic text information to obtain semantic text information about said text documents and for building said semantic text information into first semantic text networks, wherein said processing is performed by a computer;

an ontology that is linked to said first semantic text networks, wherein said ontology includes concepts and slots, wherein each of the concepts defines an entity, and wherein each of the slots defines a binary relation between two concepts; and

means for matching said first semantic text networks to second semantic text networks and for specifying meanings of said text passages to find said text passages in said text documents.