Semantic processor for recognition of whole-part relations in natural language documents

• US 8,799,776 B2
• Filed: 03/15/2007
• Issued: 08/05/2014
• Est. Priority Date: 07/31/2001
• Status: Active Grant

• Current Assignee: Allium US Holding LLC
• Original Assignee: Invention Machine Corporation (S&P Global Inc.)
• Inventors: Todhunter, James, Sovpel, Igor, Pastanohau, Dzianis, Vorontsov, Alexander, Vertel, Alexei
• Primary Examiner(s): TAPP, AMELIA L
• Application Number:
  US11/686,660
• Publication Number:
  US 20070156393A1
• Time in Patent Office: 2,700 Days
• Field of Search: 715/256
• US Class Current: 715/256
• CPC Class Codes: G06F 16/345 Summarisation for human users
• CPC Class Codes: G06F 40/10 Text processing natural language analysis G06F40/20; semantic analysis G06F40/30; processing or translation of natural language G06F40/40
• CPC Class Codes: G06F 40/20 Natural language analysis semantic analysis of natural language G06F40/30
• CPC Class Codes: G06F 40/30 Semantic analysis

Abstract

A semantic processor and method for automatically recognizing Whole-Part relations in at least one natural language electronic or digital document recognizes one or more expanded Subject-Action-Object (eSAO) sets in text, wherein each eSAO set has one or more eSAO components; matches the one or more eSAO sets against Whole-Part relationship patterns, and generates one or more eSAO Whole-Part relations based on the matching, wherein the eSAO Whole-Part relation comprises a Whole eSAO and an Part eSAO.

Claims

1. A processor-implemented method for automatically recognizing Whole-Part relations in a natural language document, the method comprising:

providing a database having a plurality of stored generic Whole-Part relationship patterns as models for determining Whole-Part relations of the outside world in external natural language documents, wherein each generic Whole-Part relationship pattern identifies;

a sense of one or more components of an expanded Subject-Action-Object (eSAO) and whether or not other components of the eSAO are non-empty; and

which components of the eSAO act as a Whole eSAO and which components act as a Part eSAO in a Whole-Part relation;

providing text from the natural language document;

recognizing one or more eSAO sets in the text, wherein each eSAO set has one or more eSAO components;

matching the one or more eSAO sets against the generic Whole-Part relationship patterns stored in the database, including determining if the eSAO component sense and non-empty requirements of a generic Whole-Part relationship pattern are satisfied by an eSAO set and, if the eSAO component sense and non-empty requirements of a generic Whole-Part relationship pattern are satisfied by an eSAO set, determining which eSAO component of the eSAO set is a Whole eSAO and which is a Part eSAO; and

generating one or more eSAO Whole-Part relations based on the matching, wherein the eSAO Whole-Part relation comprises a Whole eSAO and an Part eSAO.

2. The method according to claim 1, wherein the one or more eSAO components are one or more elements from a group comprising:

subjects, objects, actions, adjectives, prepositions, indirect objects, and adverbs.

3. The method according to claim 1, wherein:

the Whole eSAO comprises one or more of the eSAO components or a part of a single eSAO component of the one or more eSAO sets; and

the Part eSAO comprises one or more of the eSAO components or a part of a single eSAO component of the one or more eSAO sets.

4. The method according to claim 2, wherein the eSAO Whole-Part relations comprise a sequential operator relating the eSAO components of the Whole eSAO to the eSAO components of the Part eSAO, the operator including one or more of a lexical, grammatical, and semantic language indicator.

5. The method according to claim 1, further comprising:

applying parts-of-speech tags to at least portions of the text to generate tagged portions of the text; and

parsing the tagged portions of the text to generate parsed and tagged portions of the text, wherein recognizing the eSAO sets in the text is performed on the parsed and tagged portions of the text.

6. The method according to claim 5, wherein applying parts-of-speech tags is performed on preformatted portions of the text, whereby the preformatted portions of the text comprise the text with non-natural language symbols removed.

7. The method according to claim 1, wherein:

matching the one or more eSAO sets against generic Whole-Part relationship patterns comprises matching a single eSAO set; and

generating one or more eSAO Whole-Part relations based on the matching comprises generating a single eSAO Whole-Part relation.

8. The method according to claim 1, wherein:

matching the one or more eSAO sets against generic Whole-Part relationship patterns comprises matching a pair of eSAO sets; and

generating one or more eSAO Whole-Part relations based on the matching comprises generating a single eSAO Whole-Part relation based on matching the pair of eSAO sets.

9. The method according to claim 1, wherein matching the one or more eSAO sets against generic Whole-Part relationship patterns comprises accessing a generic Whole-Part pattern database that is generated by a method comprising:

recognizing eSAO sets in a corpus of text documents that does not include the natural language document;

generating a corpus of sentences, wherein each sentence contains at least one of the recognized eSAO sets;

recognizing particular cases of Whole-Part relations in the sentences;

generalizing the particular cases of Whole-Part relations into eSAO Whole-Part patterns; and

storing the eSAO Whole-Part patterns in the Whole-Part pattern database.

10. The method according to claim 1, wherein recognizing one or more expanded Subject-Action-Object (eSAO) sets in the text comprises accessing a linguistic knowledge base having a database defining eSAO components.

11. The method according to claim 1, wherein recognizing one or more expanded Subject-Action-Object (eSAO) sets in the text comprises recognizing one or more of subjects, objects, actions, adjectives, prepositions, indirect objects, and adverbs in one or more sentences of the text.

12. A computer-based method for generating a Whole-Part knowledge base by automatically recognizing Whole-Part relations in natural language documents, the method comprising:

providing a database having a plurality of stored generic Whole-Part relationship patterns as models for determining Whole-Part relations of the outside world in external natural language documents, wherein each generic Whole-Part relationship pattern identifies;

a sense of one or more components of an expanded Subject-Action-Object (eSAO) and whether or not at least two other components of the eSAO are non-empty; and

which components of the eSAO act as a Whole eSAO and which components act as a Part eSAO in a Whole-Part relation;

providing text from at least one natural language document;

recognizing one or more eSAO sets in the text, wherein each eSAO set has one or more eSAO components;

matching the one or more eSAO sets against generic Whole-Part relationship patterns comprising Whole-Part indicators, including determining if the eSAO component sense and non-empty requirements of a generic Whole-Part relationship pattern are satisfied by an eSAO set and, if the eSAO component sense and non-empty requirements of a generic Whole-Part relationship pattern are satisfied by an eSAO set, determining which eSAO component is a Whole eSAO and which is a Part eSAO;

generating one or more eSAO Whole-Part relations based on the matching, wherein the eSAO Whole-Part relation comprises a Whole eSAO and an Part eSAO; and

storing the one or more eSAO Whole-Part relations in the Whole-Part knowledge base.

13. A computer program product comprising a computer-readable medium having computer-executable instructions for performing a method for recognizing Whole-Part relations in natural language documents, the method comprising:

accessing a database having a plurality of stored generic Whole-Part relationship patterns as models for determining Whole-Part relations of the outside world in external natural language documents, wherein each generic Whole-Part relationship pattern identifies;

a sense of one or more components of an expanded Subject-Action-Object (eSAO) and whether or not other components of the eSAO are non-empty; and

which components of the eSAO act as a Whole eSAO and which components act as a Part eSAO in a Whole-Part relation;

providing text from at least one natural language document;

recognizing one or more eSAO sets in the text, wherein each eSAO set has one or more eSAO components;

matching the one or more eSAO sets against generic Whole-Part relationship patterns comprising whole-part indicators, including determining if the eSAO component sense and non-empty requirements of a generic Whole-Part relationship pattern are satisfied by an eSAO set and, if the eSAO component sense and non-empty requirements of a generic Whole-Part relationship pattern are satisfied by an eSAO set, determining which eSAO component of the eSAO set is a Whole eSAO and which is a Part eSAO;

generating one or more eSAO Whole-Part relations based on the matching, wherein the eSAO Whole-Part relation comprises a Whole eSAO and an Part eSAO.

14. The computer program product of claim 13, wherein the method further comprises storing the one or more eSAO Whole-Part relations in a Whole-Part knowledge base.

15. A semantic processor for automatically recognizing Whole-Part relations in text in electronic or digital form, the semantic processor comprising:

a linguistic knowledge base having a plurality of stored generic Whole-Part relationship patterns as models for determining Whole-Part relations of the outside world in external natural language documents, wherein each generic Whole-Part relationship pattern identifies;

a sense of one or more components of an expanded Subject-Action-Object (eSAO) and whether or not other components of the eSAO are non-empty; and

which components of the eSAO act as a Whole eSAO and which components act as a Part eSAO in a Whole-Part relation; and

a semantic analyzer comprising;

an expanded subject-action-object (eSAO) recognizer for producing one or more eSAO sets based on the text, wherein the eSAO sets are based on eSAO definitions stored in the linguistic knowledge base; and

a Whole-Part recognizer configured to match the one or more eSAO sets with known generic Whole-Part relationship patterns comprising Whole-Part indicators stored in the linguistic knowledge base, including determining if the eSAO component sense and non-empty requirements of a generic Whole-Part relationship pattern are satisfied by an eSAO set and, if the eSAO component sense and non-empty requirements of a generic Whole-Part relationship pattern are satisfied by an eSAO set, determining which eSAO component of the eSAO set is a Whole eSAO and which is a Part eSAO and to produce one or more eSAO Whole-Part relations based on the match.

16. The semantic processor of claim 15, wherein the semantic processor comprises a linguistic analyzer comprising the semantic analyzer, the linguistic analyzer further comprising:

a part-of-speech tagger configured to apply parts of speech tags to at least portions of the text; and

a parser configured to parse the text tagged by the parts-of-speech tagger and to provide the parsed and tagged text to the expanded subject-action-object (eSAO) recognizer.

17. The semantic processor of claim 16, further comprising:

a preformatter configured to receive the text in electronic or digital format and to produce preformatted text based on data stored in the linguistic knowledge base, for input to the part-of-speech tagger; and

a knowledge base generator configured to produce a Whole-Part knowledge base from the one or more eSAO Whole-Part relations generated by the linguistic analyzer.

18. The semantic processor of claim 17, wherein the preformatter is configured to perform at least one of the following functions:

remove symbols in a digital or electronic representation of the text that do not form a part of natural language text;

detect and correct mismatches or mistakes in the text; and

partition the text into structures of sentences and words.

19. The semantic processor of claim 15, further comprising a Whole-Part relationship generator configured to generate and store the generic Whole-Part relationship patterns, the Whole-Part relationship generator comprising:

a corpus linguistic analyzer configured to recognize eSAO sets in a corpus of text documents;

a corpus eSAO generator configured to generate a corpus of sentences, wherein each sentence contains at least one of the recognized eSAO sets;

a relation recognizer configured to recognize particular cases of Whole-Part relations in the sentences;

a pattern generator configured to generalize the particular cases of Whole-Part relations to eSAO Whole-Part patterns; and

a pattern tester configured to store the eSAO Whole-Part patterns in a Whole-Part pattern database.

20. The semantic processor of claim 15, wherein the one or more eSAO Whole-Part relations each comprises a Whole eSAO, a Part eSAO, and at least one sequential operator relating the Whole eSAO to the Part eSAO.

21. The semantic processor of claim 20, wherein each eSAO set based on the text comprises eSAO components and the Whole eSAO comprises one or more of the eSAO components and the Part eSAO comprises one or more of the eSAO components different than the one or more eSAO components of the Whole eSAO.

22. The semantic processor according to claim 21, wherein the eSAO components are one or more elements from a group comprising:

subjects, objects, actions, adjectives, prepositions, indirect objects, and adverbs.

23. The semantic processor according to claim 15, wherein the Whole-Part recognizer is further configured to match a single eSAO set with a generic Whole-Part relationship pattern to generate a single eSAO Whole-Part relation.

24. The semantic processor according to claim 15, wherein the Whole-Part recognizer is further configured to match a pair of eSAO sets with a generic Whole-Part relationship pattern to generate a single eSAO Whole-Part relation.