Query phrasification

• US 8,600,975 B1
• Filed: 04/09/2012
• Issued: 12/03/2013
• Est. Priority Date: 03/30/2007
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Cao, Pei, Mazumdar, Soham
• Primary Examiner(s): Singh, Amresh
• Application Number:
  US13/442,317
• Time in Patent Office: 603 Days
• Field of Search: 707/713
• US Class Current: 707/713
• CPC Class Codes: G06F 16/313 Selection or weighting of terms for indexing
• CPC Class Codes: G06F 16/3344 using natural language analysis
• CPC Class Codes: G06F 40/289 Phrasal analysis, e.g. finite state techniques or chunking

Abstract

An information retrieval system uses phrases to index, retrieve, organize and describe documents. Phrases are extracted from the document collection. Documents are the indexed according to their included phrases, using phrase posting lists. The phrase posting lists are stored in an cluster of index servers. The phrase posting lists can be tiered into groups, and sharded into partitions. Phrases in a query are identified based on possible phrasifications. A query schedule based on the phrases is created from the phrases, and then optimized to reduce query processing and communication costs. The execution of the query schedule is managed to further reduce or eliminate query processing operations at various ones of the index servers.

Claims

1. A computer-implemented method comprising:

decomposing, by at least one processor of a computer system, a search query that includes three or more words into a plurality of candidate phrasifications, including different groupings of words of the search query, each candidate phrasification comprising a disjoint union of component phrases, and each component phrase including at least one word or related word of the search query;

scoring, by at least one of the processors of the computer system, at least two of the candidate phrasifications, wherein the candidate phrasifications include one or more component phrases, and wherein the scoring is based on a probability of occurrence of each of the candidate phrasification'"'"'s component phrases, and is based on the number of component phrases constituting the candidate phrasification, wherein candidate phrasifications having relatively fewer component phrases are weighted higher than candidate phrasifications having relatively more component phrases;

selecting, by at least one of the processors of the computer system and based on scores of the candidate phrasifications, a subset of the candidate phrasification; and

executing a query of a document indexing, by at least one of the processors of the computer system, using the selected subset of candidate phrasifications, wherein the query comprises the component phrases of each selected phrasification.

2. The method of claim 1, wherein decomposing the search query comprises:

(i) restructuring a word tree derived from the search query, where the word tree comprises leaf nodes that are words and non-leaf nodes that are Boolean operators AND, OR, or NOT; and

(ii) generating a plurality of phrasifications based on the words as ordered combinations of leaf nodes.

3. The method of claim 1, wherein scoring each candidate phrasification further comprises:

applying the probability of occurrence of each component phrase, and the number of component phrases in the candidate phrasification to a function that includes at least one parameter for adjusting precision and recall of the candidate phrasification.

4. The method of claim 1, wherein scoring each phrasification comprises:

applying a scoring model;

5. The method of claim 1, wherein the probability of occurrence of the component phrases of a candidate phrasification is based on a proportion of documents to be searched that contain the component phrase.

6. The method of claim 1, wherein decomposing the search query into a candidate phrasification includes grouping and rearranging words of the search query.

7. The method of claim 1, wherein decomposing the search query includes replacing one or more of the input words with a related word.

8. The method of claim 1, wherein decomposing the search query includes replacing one or more of the component phrases with a related phrase.

9. The method of claim 8, wherein the related phrase includes a phrase synonymous with one of the component phrases.

10. The method of claim 8, wherein the related phrase includes a spell-corrected version of one of the component phrases.

11. The method of claim 1, wherein related words of the search query include synonyms of the words of the search query.

12. The method of claim 1, wherein the probability of occurrence of each component phrase in a candidate phrasification is based on an incidence of the component phrase in an indexed document collection.

13. The method of claim 1, further comprising:

executing the search query by using the selected phrasifications to identify documents containing component phrases of the selected phrasifications.

14. The method of claim 1, wherein selecting at least one candidate phrasification includes selecting a predetermined number of candidate phrasifications.

15. The method of claim 1, wherein selecting at least one candidate phrasification includes selecting a predetermined percentage of the top scoring candidate phrasifications.

16. The method of claim 1, wherein selecting at least one candidate phrasification includes selecting a number of candidate phrasifications, wherein the number is based on a load on a system that scores the candidate phrasifications or that selects the candidate phrasifications.

17. The method of claim 1, wherein selecting at least one candidate phrasification includes selecting a number of candidate phrasifications, wherein the number is based on a source of the query.

18. The method of claim 1, wherein the selected subset of candidate phrasifications is organized as a Boolean phrase tree.

19. The method of claim 1, wherein the phrase tree is rooted by an OR node.

20. The method of claim 1, wherein each of the selected candidate phrasifications are disjuncts of a root node of a Boolean phrase tree that includes the selected subset of candidate phrasifications.

21. A system comprising:

a computer program product stored on a tangible computer readable medium and comprising instructions that when executed cause a computer system to;

decompose a search query that includes three or more words into a plurality of candidate phrasifications, including different groupings of words of the search query, each candidate phrasification comprising a disjoint union of component phrases, and each component phrase including at least one word or related word of the search query;

score at least two of the candidate phrasifications, wherein the candidate phrasifications include one or more component phrases, and wherein the scoring is based on a probability of occurrence of each of the candidate phrasification'"'"'s component phrases, and is based on the number of component phrases constituting the candidate phrasification, wherein candidate phrasifications having relatively fewer component phrases are weighted higher than candidate phrasifications having relatively more component phrases;

select, based on the scores of the candidate phrasifications, at least one candidate phrasification,

select, based on the scores of the candidate phrasifications, a subset of the candidate phrasifications; and

execute a query of a document indexing using the selected subset of candidate phrasifications, wherein the query comprises the component phrases of each selected phrasification; and

one or more processors configured for executing the instructions.

22. The system of claim 21, wherein decomposing the search query comprises:

(i) restructuring a word tree derived from the search query, where the word tree comprises leaf nodes that are words and non-leaf nodes that are Boolean operators AND, OR, or NOT; and

(ii) generating a plurality of phrasifications based on the words as ordered combinations of leaf nodes.

23. The system of claim 21, wherein scoring each phrasification comprises:

applying a scoring model;

24. The system of claim 21, wherein the probability of occurrence of the component phrases of a candidate phrasification is based on an estimated probability figure.

25. The system of claim 21, wherein the probability of occurrence of the component phrases of a candidate phrasification is based on a proportion of documents to be searched that contain the component phrase.

26. The system of claim 21, wherein decomposing the search query into a candidate phrasification includes grouping and rearranging words of the search query.

27. The system of claim 21, wherein decomposing the search query includes replacing one or more of the input words with a related word.

28. The system of claim 21, wherein decomposing the search query includes replacing one or more of the component phrases with a related phrase.

29. The system of claim 28, wherein a related phrase includes a phrase synonymous with one of the component phrases.

30. The system of claim 28, wherein a related phrase includes a spell-corrected version of one of the component phrases.

31. The system of claim 21, wherein the probability of occurrence of each component phrase in a candidate phrasification is based on an incidence of the component phrase in an indexed document collection.

32. The system of claim 21, wherein related words of the search query include synonyms of the words of the search query.

33. The system of claim 21, wherein the instructions further comprise instructions to execute the search query by using the selected phrasifications to identify documents containing component phrases of the selected phrasifications.

34. The system of claim 21, wherein selecting at least one candidate phrasification includes selecting a predetermined number of candidate phrasifications.

35. The system of claim 21, wherein selecting at least one candidate phrasification includes selecting a predetermined percentage of the top scoring candidate phrasifications.

36. The system of claim 21, wherein selecting at least one candidate phrasification includes selecting a number of candidate phrasifications, wherein the number is based on a load on the system.

37. The system of claim 21, wherein selecting at least one candidate phrasification includes selecting a number of candidate phrasifications, wherein the number is based on a source of the query.

38. The system of claim 21, wherein the selected subset of candidate phrasifications is organized as a Boolean phrase tree.

39. The system of claim 21, wherein the phrase tree is rooted by an OR node.

40. The system of claim 21, wherein each of the selected candidate phrasifications are disjuncts of a root node of a Boolean phrase tree that includes the selected subset of candidate phrasifications.