Learning facts from semi-structured text

• US 7,769,579 B2
• Filed: 05/31/2005
• Issued: 08/03/2010
• Est. Priority Date: 05/31/2005
• Status: Expired due to Fees

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Zhao, Shubin, Betz, Jonathan T.
• Primary Examiner(s): Hudspeth; David R
• Assistant Examiner(s): Serrou; Abdelali
• Application Number:
  US11/142,853
• Publication Number:
  US 20060293879A1
• Time in Patent Office: 1,890 Days
• Field of Search: 704 7- 10, 704/5, 704/2, 704/51, 704/52, 707/962, 707/E17.118
• US Class Current: 704/7
• CPC Class Codes: G06F 16/353 into predefined classes
• CPC Class Codes: G06F 16/84 Mapping; Conversion
• CPC Class Codes: G06F 16/93 Document management systems
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: Y10S 707/962 Entity-attribute-value

Abstract

A method and system of learning, or bootstrapping, facts from semi-structured text is described. Starting with a set of seed facts associated with an object, documents associated with the object are identified. The identified documents are checked to determine if each has at least a first predefined number of seed facts. If a document does have at least a first predefined number of seed facts, a contextual pattern associated with the seed facts is identified and other instances of content in the document matching the contextual pattern are identified. If the document includes at least a second predefined number of the other instances of content matching the contextual pattern, then facts may be extracted from the other instances.

Claims

1. A computer-implemented method of learning facts, comprising:

at a computer system including one or more processors and memory storing one or more programs, the one or more processors executing the one or more programs to perform the operations of;

accessing an object within a fact repository, wherein the object includes a name and one or more seed facts;

identifying a set of documents having content and associated with the object name, each document in the set having at least a first predefined number of distinct seed facts in common with the seed facts of the object;

for each of the documents in the identified set;

identifying in the document a contextual pattern associated with the respective seed facts in the document;

confirming that the document includes at least a second predefined number of instances of content matching the contextual pattern in addition to the respective seed facts; and

only when the confirming is successful, extracting an extracted fact from a respective instance of content matching the contextual pattern and merging the extracted fact into the object;

wherein the first predefined number is greater than one and the second predefined number is greater than one.

2. The method of claim 1, further comprising repeating the extracting and merging operations for one or more instances of content matching the contextual pattern in the document.

3. The method of claim 1, wherein the extracted and merged fact comprises an attribute-value pair that is distinct from all other attribute-value pairs of the object.

4. The method of claim 1, further comprising:

identifying an attribute-value pair in the document that matches a respective attribute-value pair of the object; and

adding an identifier of the document to a list of documents associated with the respective attribute-value pair of the object.

5. The method of claim 4, further comprising generating a confidence value for each attribute-value pair of the object that is based on the documents in the list of documents associated with the attribute-value pair.

6. The method of claim 4, further comprising generating a confidence value for each attribute-value pair of the object that corresponds to a number of documents in the list of documents associated with the attribute-value pair.

7. A system for learning facts, comprising:

one or more processors; and

memory storing one or more modules having instructions for execution by the one or more processors, including instructions;

to access an object within a fact repository, wherein the object includes a name and one or more seed facts;

to identify a set of documents having content and associated with the object name, each document in the set having at least a first predefined number of distinct seed facts in common with the seed facts of the object;

for each of the documents in the identified set;

to identify in the document a contextual pattern associated with the respective seed facts in the document; and

to confirm that the document includes at least a second predefined number of instances of content matching the contextual pattern in addition to the respective seed facts; and

to extract an extracted fact only from a respective instance of content matching the contextual pattern and merge the extracted fact into the object;

wherein the first predefined number is greater than one and the second predefined number is greater than one.

8. The system of claim 7, wherein the one or more modules include instructions to repeatedly extract and merge attribute-value pairs from instances of content matching the contextual pattern in the document.

9. The system of claim 7, wherein the extracted and merged fact comprises an attribute-value pair that is distinct from all other attribute-value pairs of the object.

10. The system of claim 7, wherein the one or more modules include instructions to:

identify an attribute-value pair in the document that matches a respective attribute-value pair of the object; and

add an identifier of the document to a list of documents associated with the respective attribute-value pair of the object.

11. The system of claim 10, further comprising instructions to generate a confidence value for each attribute-value pair of the object that is based on the documents in the list of documents associated with the attribute-value pair.

12. The system of claim 10, further comprising instructions to generate a confidence value for each attribute-value pair of the object that corresponds to a number of documents in the list of documents associated with the attribute-value pair.

13. A computer readable storage medium storing one or more programs for execution by a computer system, the one or more programs comprising instructions for:

accessing an object within a fact repository, wherein the object includes a name and one or more seed facts;

identifying a set of documents having content and associated with the object name, each document in the set having at least a first predefined number of distinct seed facts in common with the seed facts of the object;

for each of the documents in the identified set;

identifying in the document a contextual pattern associated with the respective seed facts in the document;

confirming that the document includes at least a second predefined number of instances of content matching the contextual pattern in addition to the respective seed facts; and

only when the confirming is successful, extracting an extracted fact from a respective instance of content matching the contextual pattern and merging the extracted fact into the object;

wherein the first predefined number is greater than one and the second predefined number is greater than one.

14. The computer readable storage medium of claim 13, further comprising repeating the extracting and merging operations for one or more instances of content matching the contextual pattern in the document.

15. The computer readable storage medium of claim 13, wherein the extracted and merged fact comprises an attribute-value pair that is distinct from all other attribute-value pairs of the object.

16. The computer readable storage medium of claim 13, further comprising instructions for:

identifying an attribute-value pair in the document that matches a respective attribute-value pair of the object; and

adding an identifier of the document to a list of documents associated with the respective attribute-value pair of the object.

17. The computer readable storage medium of claim 16, further comprising instructions for generating a confidence value for each attribute-value pair of the object that is based on the documents in the list of documents associated with the attribute-value pair.

18. The computer readable storage medium of claim 16, further comprising instructions for generating a confidence value for each attribute-value pair of the object that corresponds to a number of documents in the list of documents associated with the attribute-value pair.

19. A system for learning facts, comprising:

one or more processors; and

memory storing one or more modules having instructions for execution by the one or more processors;

means for accessing an object within a fact repository, wherein the object includes a name and one or more seed facts;

means for identifying a set of documents having content and associated with the object name, each document in the set having at least a first predefined number of distinct seed facts in common with the seed facts of the object;

means, for each of the documents in the identified set;

for identifying in the document a contextual pattern associated with the respective seed fact in the document;

for confirming that the document includes at least a second predefined number of instances of content matching the contextual pattern in addition to the respective seed facts; and

only when the confirming is successful, for extracting an extracted fact from a respective instance of content matching the contextual pattern and merging the extracted fact into the object;

wherein the first predefined number is greater than one and the second predefined number is greater than one.