INDEXING ROLE HIERARCHIES FOR WORDS IN A SEARCH INDEX

• US 20090063473A1
• Filed: 08/29/2008
• Published: 03/05/2009
• Est. Priority Date: 08/31/2007
• Status: Active Grant

• Current Assignee: Zhigu Holdings Limited
• Original Assignee: Powerset Inc (Microsoft Corporation)
• Inventors: THIONE, Giovanni L., VAN DEN BERG, Martin HENK, CROUCH, Richard S., WALTERS, Chad P.
• Granted Patent: US 8,229,730 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/3344 using natural language analysis
• CPC Class Codes: G06F 40/30 Semantic analysis

Abstract

Methods, systems and computer readable media for finding documents in a data store that match a natural language query submitted by a user are provided. The documents and queries are matched by determining that words within the query have the same relationship to each other as the same words in the document. Documents are semantically analyzed and words in the document are indexed along with the role the word plays in a sentence. The initial semantic role may be generalized using a role hierarchy and stored in the index along with the original role. A similar analysis may be used with the search query to find words used in the same role in both the query and the document.

Claims

1. One or more computer-readable media having computer-executable instructions embodied thereon for performing a method of expanding and indexing a semantic relationships between words within a sentence, the method comprising:

receiving a first electronic document having content that includes text;

breaking the text into one or more sentences;

determining a first role for a word in the sentence;

determining that the word has more than one potential role;

assigning a second role to the word; and

storing the word with the first role and the second role in an index that associates the word and the sentence with the first electronic document, thereby allowing a search engine to match the first electronic document with a search query that includes similar words having similar roles.

2. The media of claim 1, wherein the method further comprises:

receiving the search query containing the word;

determining that the word is used in the first role within the search query;

determining that the word has more than one potential role within the search query; and

assigning a third role to the word.

3. The media of claim 2, wherein the method further comprises:

searching the index to find one or more electronic documents having the word used in one or more of the first role and the third role.

determining the first electronic document is a first match to the search query;

determining a second electronic document having the word assigned to the first role and the third role is a second match to the search query; and

ranking the second electronic document as a better match to the search query than the first electronic document because the second electronic document has the word in the same roles that are assigned to the word in the search query.

4. The media of claim 2, wherein the method further comprises:

determining that at least two document associated with entries in the index contain the word associated with the first role or the third role; and

assigning a higher ranking to documents that contain the word associated with more of the same roles assigned to the word in the search query.

5. The media of claim 1, wherein the second role is assigned using a role hierarchy, wherein the role hierarchy includes a hierarchy of roles including a root role that is general and at least one level of sub roles that are more specific than the root role.

6. The media of claim 1, wherein storing the word with the first role and the second role includes creating a first term in the index including the word and the first role and a second term in the index including the word and the second role.

7. The media of claim 1, wherein storing the word with the first role and the second role includes creating a term in the index including the word and the first role and the second role.

8. A computerized system for matching a search query to text within a one or more documents comprising:

a first semantic interpretation component for

(1) assigning a primary role to words in the search query;

(2) assigning an additional role to at least one word from the search query having the primary role using a role hierarchy;

a matching component for determining that at least one document referenced in an index includes a word from the search query, wherein the word in the at least one document and the word in the search query are assigned at least one of the same roles.

9. The computerized system of claim 8 further comprising a presentation component for presenting information describing the at least one document.

10. The computerized system of claim 8, wherein the first semantic interpretation component generates additional words to search the index by adding at least one of a synonym and a hypernym for one or more words in the search query to a list of search terms, wherein a search term includes a single word and an associated role, and wherein the synonym and the hypernym form search terms with the roles assigned to the one or more words for which the synonym and the hypernym are synonyms and hypernyms.

11. The computerized system of claim 8, further comprising a ranking component for ranking the one or more documents, wherein a higher ranking is given when the word in the search query that is assigned more than one role has multiple roles in common with the word occurring in the index.

12. The computerized system of claim 8, further comprising:

a second semantic interpretation component for

(1) receiving a sentence that is part of a document;

(2) assigning a primary role for words in the sentence;

(3) assigning at least one additional role to at least one word in the sentence having the primary role using the role hierarchy; and

an indexing component for storing the words in the sentence with the at least one role assigned to the words in the sentence in the index that associates the words in the sentence and the sentence with the document.

13. The computerized system of claim 12, wherein the indexing component indexes the words and the roles in a term that includes a single word and an assigned role, wherein each term includes one word and one role.

14. The computerized system of claim 12, wherein the indexing component indexes the words and the roles in a term that includes a single word and an assigned role, wherein each term includes one word and at least one role, thereby allowing multiple roles to be include in a single term.

15. One or more computer-storage media having computer-executable instructions embodied thereon for performing a method of indexing document contents using a role hierarchy, the method comprising:

receiving a sentence that is part of a document;

assigning a role to words in the sentence;

assigning at least one additional role to the words using the role hierarchy, wherein the role hierarchy includes a hierarchy of roles including a root role that is general and at least one level of sub roles that are more specific than the root role, thereby assigning at least one word at least two roles; and

storing the words with and one or more roles assigned to individual words in an index that associates the words and the sentence with the document.

16. The media of claim 15, wherein storing the words and the roles includes generating a term that includes a word and an assigned role, wherein each term includes one word and one role.

17. The media of claim 15, wherein storing the words and the roles includes generating a term that includes a word and an assigned role, wherein each term includes one word and at least one role, thereby allowing multiple roles to be include in a single term.

18. The media of claim 15, wherein the method further comprises:

receiving a natural language search query containing a plurality of words;

assigning a primary role to the words in the plurality of words;

assigning at least one additional role to at least one word in the plurality of words using the role hierarchy; and

determining that one or more documents referenced in the index have at least one word in the plurality of words assigned one or more of the same roles.

19. The media of claim 18, wherein the method further comprises:

ranking the one or more documents, wherein documents with more matching words and the roles are given a higher rank, and wherein more ranking points are given to a word in the plurality of words that is associated with multiple roles has more roles in common with an occurrence of the word in the index.

20. The media of claim 19, wherein the method further includes displaying the one or more documents according to a rank assigned to each of the one or more documents.