Semantic processor for recognition of cause-effect relations in natural language documents

• US 9,009,590 B2
• Filed: 10/24/2005
• Issued: 04/14/2015
• Est. Priority Date: 07/31/2001
• Status: Active Grant

• Current Assignee: Allium US Holding LLC
• Original Assignee: Invention Mine LLC
• Inventors: Todhunter, James, Sovpel, Igor, Pastanohau, Dzianis, Vorontsov, Alexander
• Primary Examiner(s): TAPP, AMELIA L
• Application Number:
  US11/258,250
• Publication Number:
  US 20060041424A1
• Time in Patent Office: 3,459 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 40/30 Semantic analysis

Abstract

A Semantic Processor for the recognition of Cause-Effect relations in natural language documents which includes a Text Preformatter, a Linguistic Analyzer and a Cause-Effect Knowledge Base Generator. The Semantic Processor provides automatic recognition of cause-effect relation both inside single fact and between the facts in arbitrary text documents, where the facts are also automatically extracted from the text in the form of seven-field semantic units. The recognition of Cause-Effect relations is carried out on the basis of linguistic (including semantic) text analysis and a number of recognizing linguistic models built in the form of patterns.

Claims

1. A computer-implemented method for recognizing cause-effect relations in natural language documents to generate a cause-effect knowledge base comprising:

recognizing a first expanded Subject-Action-Object (eSAO) set in a text from at least one natural language document using a processor, the first eSAO set including two or more eSAO components in the text grouped together to form a first semantic unit;

recognizing a second eSAO set in the text from the at least one natural language document using the processor, the second eSAO set including two or more eSAO components in the text grouped together to form a second semantic unit, the two or more eSAO components grouped together to form the second semantic unit being different from the two or more eSAO components grouped together to form the first semantic unit;

matching the first eSAO set and the second eSAO set against cause-effect (C-E) relationship patterns stored in a memory, wherein the C-E relationship patterns include indicators of a presence of a C-E relation in a sentence and information about an eSAO set in the sentence that acts as a cause part of the C-E relation and an eSAO set in the sentence that acts as an effect part of the C-E relation;

generating one or more eSAO cause-effect relations based on the matching, wherein each eSAO cause-effect relation identifies one of the first eSAO set and the second eSAO set as a cause eSAO and the other of the first eSAO set and the second eSAO set as an effect eSAO; and

storing the one or more eSAO cause-effect relations in a knowledge base.

2. The method according to claim 1, wherein the first eSAO set includes three or more eSAO components in the text grouped together to form the first semantic unit, and wherein each of the three or more eSAO components in the text comprise text related to one or more elements of the group consisting of:

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

3. The method according to claim 1, wherein the cause eSAO comprises one or more of the eSAO components of the first eSAO set and the effect eSAO comprises one or more eSAO components of the second eSAO set.

4. The method according to claim 3, wherein the one or more eSAO cause-effect relations each comprise a sequential operator relating the eSAO components of the cause eSAO to the eSAO components of the effect eSAO with lexical, grammatical, and/or semantic language means.

5. The method according to claim 1, wherein matching the first eSAO set and the second eSAO set against cause-effect relationship patterns comprises accessing a cause-effect pattern database, and wherein the cause-effect pattern database is generated by a method comprising:

recognizing eSAO sets in a corpus of text documents;

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

recognizing particular cases of cause-effect relations in the sentences;

generalizing the particular cases of cause-effect relations to eSAO cause-effect patterns;

storing the eSAO cause-effect patterns in a cause-effect pattern database.

6. The method according to claim 1, wherein recognizing the first expanded Subject-Action-Object (eSAO) set in the text comprises accessing a linguistic knowledge base having a database defining eSAO components.

7. The computer implemented method of claim 1, wherein the two or more eSAO components in the text grouped together to form the first semantic unit are three or more eSAO components that include a subject component, an action component, and an object component, and matching the first eSAO set and the second eSAO set against cause-effect (C-E) relationship patterns stored in the memory includes identifying a cause-effect relationship pattern using the action component of the first eSAO set.

8. The semantic processor of claim 1, wherein the first eSAO set includes four or more eSAO components in the text grouped together to form the first semantic unit.

9. The semantic processor of claim 8, wherein the second eSAO set includes four or more eSAO components in the text grouped together to form the second semantic unit.

10. A computer-readable medium having computer-executable instructions for performing a method for recognizing cause-effect relations in natural language documents to generate a cause-effect knowledge base, the method comprising:

providing text from at least one natural language document;

recognizing a first expanded Subject-Action-Object (eSAO) set that includes two or more eSAO components in the text grouped together to form a first semantic unit;

recognizing a second eSAO set in the text that includes two or more eSAO components in the text grouped together to form a second semantic unit, the two or more eSAO components grouped together to form the second semantic unit being different from the two or more eSAO components grouped together to form the first semantic unit;

matching the first and second eSAO sets against cause-effect (C-E) relationship patterns stored in a memory, wherein the C-E relationship patterns include indicators of a presence of a C-E relation in a sentence and information about an eSAO set in the text that acts as a cause part of the C-E relation and an eSAO set in the text acts as an effect part of the C-E relation;

generating one or more eSAO cause-effect relations based on the matching, wherein each eSAO cause-effect relation identifies one of the first eSAO set and the second eSAO set as a cause eSAO and the other of the first eSAO set and the second eSAO set as an effect eSAO; and

storing the one or more eSAO cause-effect relations in a knowledge base.

11. The computer implemented method of claim 10, wherein the cause-effect (C-E) relationship patterns stored in a memory include at least of one of the following patterns:

A.) adverbial modifier of purpose TO VB|DO|HV pattern;

B.) adverbial modifier of cause BY VBG pattern;

C.) adverbial modifier of purpose FOR VBG pattern;

D.) adverbial modifier of purpose THUS VGB pattern;

E.) subject in the form of a participle phrase pattern;

F.) adverbial modifier of cause DEPEND ON pattern;

G.) complex sentence CONDUCTION CAUSE pattern; and

H.) complex sentence CONJUNCTION EFFECT pattern.

12. A semantic processor comprising computer hardware that executes logic to recognize cause-effect relations in text in electronic or digital form the processor comprising:

a linguistic knowledge base stored in at least one storage media; and

a semantic analyzer comprising;

an expanded subject-action-object (eSAO) recognizer configured to recognize a first eSAO set and a second eSAO set in the text, the first eSAO set including two or more eSAO components and the second eSAO set including two or more eSAO components that are different than the two or more eSAO components of the first eSAO set, the first and second eSAO sets being based on eSAO definitions stored in the linguistic knowledge base; and

a cause-effect recognizer configured to match the first and second eSAO sets with known cause-effect relationship patterns stored in the linguistic knowledge base, the known cause-effect relationship patterns including indicators of a presence of a cause-effect relation in a sentence and information about an eSAO set in the sentence that acts as a cause part of the C-E relation and an eSAO set in the sentence that acts as an effect part of the cause-effect relation, and to produce one or more eSAO cause-effect relations based on the match, wherein the produced cause-effect relations each identify one of the first eSAO set and the second eSAO set as a cause eSAO and the other of the first eSAO set and the second eSAO set as an effect eSAO.

13. The semantic processor according to claim 12, wherein the semantic processor comprises a linguistic analyzer that includes:

a parts-of speech tagger for receiving preformatted text based on the text in electronic or digital format; and

a parser for receiving the text tagged by the parts-of-speech tagger and providing parsed text to the semantic analyzer,

wherein the parts-of-speech tagger and the parser operate with data stored in the linguistic knowledge base.

14. The semantic processor according to claim 13, further comprising:

a preformatter for receiving the text in electronic or digital format and producing preformatted text based on data stored in the linguistic knowledge base; and

a knowledge base generator for producing a knowledge base from the one or more eSAO cause-effect relations generated by the linguistic analyzer.

15. The semantic processor according to claim 14, wherein the preformatter performs at least one of the following functions:

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

detection and correction of any mismatches or mistakes in the text; and

partitioning the text into structures of sentences and words.

16. The semantic processor according to claim 12, wherein the known cause-effect relationship patterns are created by a cause-effect relationship generator comprising:

a corpus linguistic analyzer for recognizing eSAO sets in a corpus of text documents;

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

a relation recognizer for recognizing particular cases of cause-effect relations in the sentences;

a pattern generator for generalizing the particular cases of cause-effect relations to eSAO cause-effect patterns; and

a pattern tester for storing the eSAO cause-effect patterns in a cause-effect pattern database, wherein the cause-effect pattern database is a database within the linguistic knowledge base.

17. The semantic processor according to claim 12, wherein the eSAO cause-effect relation comprises a cause eSAO, an effect eSAO, and at least one sequential operator relating the cause eSAO to the effect eSAO.

18. The semantic processor according to claim 17, wherein the cause eSAO comprises one or more eSAO components and the effect eSAO comprises one or more eSAO components different than the one or more eSAO components of the cause eSAO.

19. The semantic processor according to claim 18, wherein the eSAO components of the cause eSAO and the effect eSAO each comprise text related to one or more elements of the group consisting of:

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

20. The semantic processor of claim 12, wherein the cause-effect recognizer is adapted to compare the first and second eSAO sets against all of the following cause-effect relationship patterns:

A.) actions with a cause sense pattern, in which an action field of the pattern has a cause sense, a subject field of the pattern has a cause label, and an object field of the pattern has an effect label;

B.) actions with a result from sense pattern, in which an action field of the pattern has a result from sense, a subject field of the pattern has an effect label, and an object field of the pattern has a cause label;

C.) cause of construction in object pattern, in which an object field of the pattern has a cause of sense and an effect label, the action field of the pattern has a be sense, and a subject field of the pattern has a cause label;

D.) cause of construction in subject pattern, in which a subject field of the pattern has a cause of sense and an effect label, the action field of the pattern has a be sense, and an object field of the pattern has a cause label;

E.) actions with technical sense pattern, in which an action field of the pattern has a technical action sense and an effect label and a subject field of the pattern has a cause field;

F.) a first process noun in object pattern, in which the object has a process noun or PROCESS_NN sense and an effect label, an action field of the pattern has a LINK ACTION_sense, and a subject field of the pattern has a cause label; and

G.) a second process noun in object pattern, in which the object has a process noun or PROCESS_NN sense, an action field of the pattern has a LINK_TECH _ACTION sense and an effect label, and a subject field of the pattern has a cause label.

21. The semantic processor of claim 12, wherein the first eSAO set includes three or more eSAO components in the text grouped together to form the distinct semantic units.

22. The semantic processor of claim 12, wherein the first eSAO set includes four or more eSAO components in the text grouped together to form the distinct semantic units.

23. A semantic processor comprising computer hardware that executes logic to recognize cause-effect relations in text in electronic or digital form the processor comprising:

a linguistic knowledge base stored in at least one storage media; and

a semantic analyzer comprising;

an expanded subject-action-object (eSAO) recognizer that produces 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 wherein each eSAO set includes three or more eSAO components in the text; and

a cause-effect recognizer that matches the one or more eSAO sets with known cause-effect relationship patterns stored in the linguistic knowledge base and produces one or more eSAO cause-effect relations based on the match, wherein the cause-effect relations each include indicators of a presence of a cause-effect relation in a sentence and information about which eSAO components or whole eSAOs in the sentence act as a cause part of the cause-effect relation and which eSAO components or whole eSAOs in the sentence act as an effect part of the cause-effect relation,

wherein the cause-effect (C-E) relationship patterns stored in a memory include at least of one of the following patterns;

A.) actions with cause sense pattern, in which an action field of the pattern has a cause sense, a subject field of the pattern has a cause label, and an object field of the pattern has an effect label;

B.) actions with result from sense pattern, in which an action field of the pattern has a result from sense, a subject field of the pattern has an effect label, and an object field of the pattern has a cause label;

C.) cause of construction in object pattern, in which an object field of the pattern has a cause of sense and an effect label, the action field of the pattern has a be sense, and a subject field of the pattern has a cause label;

D.) cause of construction in subject pattern, in which a subject field of the pattern has a cause of sense and an effect label, the action field of the pattern has a be sense, and an object field of the pattern has a cause label;

E.) actions with technical sense pattern, in which an action field of the pattern has a technical action sense and an effect label and a subject field of the pattern has a cause label;

F.) a first process noun in object pattern, in which the object has a process noun or PROCESS_NN sense and an effect label, an action field of the pattern has a LINK_ACTION sense, and a subject field of the pattern has a cause label; and

G.) a second process noun in object pattern, in which the object has a process noun or PROCESS_NN sense, an action field of the pattern has a LINK_TECH _ACTION sense and an effect label, and a subject field of the pattern has a cause label.