Method and system for extending keyword searching to syntactically and semantically annotated data

• US 7,526,425 B2
• Filed: 12/13/2004
• Issued: 04/28/2009
• Est. Priority Date: 08/14/2001
• Status: Expired due to Fees

• Current Assignee: Fiver LLC
• Original Assignee: EVRI Incorporated
• Inventors: Marchisio, Giovanni B., Pochman, Lubos, Tusk, Carsten, Liang, Jisheng, Brown, Matthew E., Dhillon, Navdeep S., Nguyen, Thien, Koperski, Krzysztof
• Primary Examiner(s): Hudspeth, David R
• Assistant Examiner(s): Sked, Matthew J
• Application Number:
  US11/012,089
• Publication Number:
  US 20050267871A1
• Time in Patent Office: 1,597 Days
• Field of Search: None
• US Class Current: 704/9
• CPC Class Codes: G06F 16/3338 Query expansion
• CPC Class Codes: Y10S 707/99931 Database or file accessing
• CPC Class Codes: Y10S 707/99943 Generating database or data structure, e.g. via user interface

Abstract

Methods and systems for extending keyword searching techniques to syntactically and semantically annotated data are provided. Example embodiments provide a Syntactic Query Engine (“SQE”) that parses, indexes, and stores a data set as an enhanced document index with document terms as well as information pertaining to the grammatical roles of the terms and ontological and other semantic information. In one embodiment, the enhanced document index is a form of term-clause index, that indexes terms and syntactic and semantic annotations at the clause level. The enhanced document index permits the use of a traditional keyword search engine to process relationship queries as well as to process standard document level keyword searches. In one embodiment, the SQE comprises a Query Processor, a Data Set Preprocessor, a Keyword Search Engine, a Data Set Indexer, an Enhanced Natural Language Parser (“ENLP”), a data set repository, and, in some embodiments, a user interface or an application programming interface.

Claims

1. A method in a computer system for preparing a corpus of documents for performing electronic searches, each document having at least one sentence, each sentence having a plurality of terms, comprising:

for each sentence of each document,

parsing the sentence under the control of the computer system to generate a parse structure having a plurality of syntactic elements that correspond to the terms of the sentence;

determining from the structure of the parse structure and the plurality of syntactic elements a corresponding grammatical role for each of a plurality of the terms of the sentence, each grammatical role being at least one of a subject, an object, a governing verb, a modifier, or a part of a prepositional phrase;

normalizing the plurality of terms of the sentence having corresponding grammatical roles to a plurality of tagged terms, each tagged term indicating an association between the term of the sentence that corresponds to the grammatical role and an associated tag type that specifies the corresponding grammatical role, wherein at least one of the tagged terms has an associated tag type that specifies that the associated term of the sentence is a subject or an object of the sentence, wherein at least one of the tagged terms has an associated tag type that specifies that the associated term of the sentence is a modifier of another term of the sentence that has an associated tag type that specifies that the another term is a subject, object, or verb of the sentence, and wherein at least one of the tagged terms has an associated tag type that additionally specifies semantic information that refers to an entity type that identifies the associated term of the sentence as a type of person, location, or thing; and

transforming each sentence to an enhanced data structure of terms stored as one or more inverted indexes of terms annotated with relationship information, wherein the plurality of the tagged terms are stored therein and indexed as additional terms of the sentence, each additional term including the term of the sentence and the associated tag type, thereby enabling a search engine to perform relationship searches by determining from the enhanced data structure whether a designated search term having an associated tag type that specifies a grammatical role and/or an entity type is present in the sentence in a same role, in a manner similar to the manner the search engine uses to determine whether a designated term is present in the sentence, at least one of the relationship searches capable of returning a plurality of relationships between at least two entities as a result of a single search specification.

2. The method of claim 1 wherein the search engine is at least one of a keyword search engine or a Boolean search engine.

3. The method of claim 1 wherein the search engine performs string matching to determine whether the designated term having the associated tag type is present in the sentence in the same role.

4. The method of claim 1 wherein the search engine performs pattern matching to determine whether the designated term having the associated tag type is present in the sentence in the same role.

5. The method of claim 1 wherein the one or more inverted indexes of terms is an augmented term-document matrix configured to be searched using pattern matching or string matching to determine sentences that are similar to a designated relationship query.

6. The method of claim 1 wherein the transforming each sentence to the enhanced data structure is performed for each clause of each sentence such that the tagged terms are treated as additional terms of each clause of the sentence and the search engine determines whether the designated syntactic term having the associated tag type is present in each clause.

7. The method of claim 1 wherein the at least one of the tagged terms that has the associated tag type that specifies semantic information specifies an action attribute tag.

8. The method of claim 1 wherein the at least one of the tagged terms that has the associated tag type that specifies semantic information specifies an ontology path.

9. The method of claim 1 wherein the normalizing the plurality of terms of the sentence having corresponding grammatical roles to the plurality of tagged items comprises applying linguistic normalization techniques to the plurality of terms of the sentence having corresponding grammatical roles to generate the plurality of tagged terms.

10. The method of claim 9 wherein the linguistic normalization techniques include applying at least one of a transformational grammar rule, a coreference resolution rule, a verbalization rule, or a verb sense rule.

11. The method of claim 10 wherein the verbalization rule is at least one of a noun verbalization rule, an adjective verbalization rule, or an adverb verbalization rule.

12. The method of claim 10 wherein the verbalization rule performs verb phrase simplification.

13. The method of claim 10 wherein the coreference resolution rule is applied to at least one of a noun, a pronoun, a noun phrase, a pronoun phrase, alias, abbreviation, or acronym.

14. The method of claim 9 wherein the linguistic normalization techniques include applying at least one rule that normalizes a set of synonyms or acronyms to a standard term or phrase.

15. The method of claim 9 wherein the linguistic normalization techniques comprise identifying and generating tagged terms that include hypernyms and hyponyms.

16. The method of claim 9 wherein the linguistic normalization techniques comprise identifying and generating tagged terms that include action attributes.

17. The method of claim 16 wherein the action attributes comprise identification of a verb tense.

18. The method of claim 16 wherein the action attributes comprise a verb mood or modality indication that specifies that the verb indicates a possibility, subjunctive, irrealis, negation, conditional, or causal relationship.

19. The method of claim 16 wherein the action attributes comprise similar verbs.

20. The method of claim 16 wherein the action attributes comprise troponyms, verb entailments, or hypernyms.

21. The method of claim 1, further comprising:

receiving a query that specifies a relationship search that designates at least one of a term and an indication of a grammatical relationship associated with the designated term or a tag type that specifies a type of grammatical role, person, location, or thing;

translating the query to a set of Boolean expressions;

executing a search engine that evaluates the Boolean expressions against the enhanced data structures of the sentences to determine a set of sentence clauses that match the query; and

returning indications to the set of matching sentence clauses.

22. The method of claim 21 wherein the received query specifies the relationship search by means of a natural language query that is transformed to the designated at least one of the term or the tag type.

23. The method of claim 21 wherein the received query specifies a relationship search in combination with a document level Boolean search for at least one keyword to be located in each matching document.

24. The method of claim 21 wherein the received query specifies a relationship search that is constrained by an expression that indicates a keyword search of the documents for at least one search term.

25. The method of claim 21 wherein the received query specifies a relationship search that is constrained by a meta-data tag expression.

26. The method of claim 21 wherein the received query specifies a relationship search that is constrained by an expression that indicates a value of a prepositional phrase.

27. The method of claim 21 wherein the tag type specifies at least one of a an entity specification, or a path in an ontology.

28. The method of claim 21 wherein the tag type specifies a subject, an object, or a verb.

29. The method of claim 21 wherein the relationship search specifies a search term using a wildcard.

30. The method of claim 29 wherein the wildcard indicates a single character, range of characters, whole word, range of words, or a specific occurrence of a word.

31. The method of claim 21 wherein the relationship search designates a value of at least one of a subject, an object, or a verb and the search engine determines all clauses in the corpus of documents where a grammatical relationship exists that satisfies the designated value of the at least one subject, object, or verb.

32. The method of claim 31 wherein the relationship search designates a value of a subject and the search engine determines a corresponding object and a corresponding verb of all clauses that contain a subject having the designated value.

33. The method of claim 31 wherein the relationship search designates a value of an object and the search engine determines a corresponding subject and a corresponding verb of all clauses that contain an object having the designated value.

34. The method of claim 31 wherein the relationship search designates a value of a verb and the search engine determines a corresponding subject and a corresponding object of all clauses that contain a verb having the designated value or a similar verb to the designated value.

35. The method of claim 31 wherein the relationship search designates a wildcard for at least one of the values of the designated at least one subject, object, or verb.

36. The method of claim 31 wherein the search engine considers the presence in a corpus sentence clause of a term used in a modifier grammatical role as a modifier of the designated value of the subject, object, or verb as a match to the designated value of the subject, object, or verb.

37. The method of claim 21 wherein the search engine is an off-the-shelf keyword search engine.

38. The method of claim 1, further comprising:

receiving a script that specifies a plurality of queries in a script language, each query specifying a relationship search that designates at least one of a term and an indication of a grammatical relationship that encompasses the term or a tag type that specifics a type of grammatical role, person, location, or thing;

translating the plurality of queries to a set of Boolean expressions;

executing a search engine that evaluates the Boolean expressions against the enhanced data structures of the sentences to determine a set of sentence clauses that match the Boolean expressions according to the script.

39. The method of claim 38 wherein the script comprises at least one of control flow instructions, group constructs, query order, or functions.

40. The method of claim 1 wherein the one or more inverted indexes of terms is a matrix that tracks occurrences of the terms across the corpus of documents.

41. The method of claim 40 wherein the matrix is at least one of a term-document matrix, a term-sentence matrix, or a term-clause matrix.

42. The method of claim 40 wherein the matrix is a plurality of term-clause matrices, each corresponding to a different grammatical role.

43. The method of claim 42 wherein the plurality of term-clause matrices comprise a subject index, an object index, and a verb index.

44. The method of claim 40 wherein each indexed term and additional term is associated with a location that corresponds to a particular clause, sentence, and document.

45. A computer-readable memory medium containing contents that, when executed, causes a computing system to index a corpus of documents for electronic searching, each document having at least one sentence, each sentence having a plurality of terms, by performing a method comprising:

for each sentence of each document,

parsing the sentence to generate a parse structure having a plurality of syntactic elements that correspond to the terms of the sentence;

determining from the structure of the parse structure and the plurality of syntactic elements a corresponding grammatical role for each of a plurality of the terms of the sentence, each grammatical role being at least one of a subject, an object, a governing verb, a modifier, or a part of a prepositional phrase;

normalizing the plurality of terms of the sentence having corresponding grammatical roles to a plurality of tagged terms, each tagged term indicating an association between the term of the sentence that corresponds to the grammatical role and an associated tag type that specifies the corresponding grammatical role, wherein at least one of the tagged terms has an associated tag type that specifies that the associated term of the sentence is a subject or an object of the sentence, wherein at least one of the tagged terms has an associated tag type that specifies that the associated term of the sentence is a modifier of another term of the sentence that has an associated tag type that specifies that the another term is a subject, object, or verb of the sentence, and wherein at least one of the tagged terms has an associated tag type that additionally specifies semantic information that refers to an entity type that identifies the associated term of the sentence as a person, location, or thing; and

transforming each sentence to an enhanced data structure of terms stored as one or more inverted indexes of terms annotated with relationship information, wherein the plurality of the tagged terms are stored therein and indexed as additional terms of the sentence, each additional term including the term of the sentence and the associated tag type, thereby enabling a search engine to perform relationship searches by determining from the enhanced data structure whether a designated search term having an associated tag type that specifies a grammatical role and/or an entity type is present in the sentence in a same role, in a manner similar to the manner the search engine uses to determine whether a designated term is present in the sentence, at least one of the relationships search capable of returning a plurality of relationships between at least two entities as a result of a single search specification.

46. The memory medium of claim 45 wherein the search engine is at least one of a keyword search engine or a Boolean search engine.

47. The memory medium of claim 45 wherein the search engine performs string matching to determine whether the designated term having the associated tag type is present in the sentence in the same role.

48. The memory medium of claim 45 wherein the search engine performs pattern matching to determine whether the designated term having the associated tag type is present in the sentence in the same role.

49. The memory medium of claim 45 wherein the one or more inverted indexes of terms is an augmented term-document matrix configured to be searched using pattern matching or string matching to determine sentences that are similar to a designated relationship query.

50. The memory medium of claim 45 wherein the transforming each sentence to the enhanced data structure is performed for each clause of each sentence such that the tagged terms are treated as additional terms of each clause of the sentence and the search engine determines whether the designated syntactic term having the associated tag type is present in each clause.

51. The memory medium of claim 45 wherein the at least one of the tagged terms that has the associated tag type that specifies semantic information specifies an action attribute tag.

52. The memory medium of claim 45 wherein the at least one of the tagged terms that has the associated tag type that specifies semantic information specifies an ontology path.

53. The memory medium of claim 45 wherein the normalizing the plurality of terms of the sentence having corresponding grammatical roles to the plurality of tagged terms comprises applying linguistic normalization techniques to the plurality of terms of the sentence having corresponding grammatical roles to generate the plurality of tagged terms.

54. The memory medium of claim 53 wherein the linguistic normalization techniques include applying at least one of a one transformational grammar rule, a coreference resolution rule, a verbalization rule, or a verb sense rule.

55. The memory medium of claim 53 wherein the linguistic normalization techniques include applying at least one rule that normalizes a set of synonyms or acronyms to a standard term or phrase.

56. The memory medium of claim 53 wherein the linguistic normalization techniques comprise identifying and generating tagged terms that include hypernyms and hyponyms.

57. The memory medium of claim 45, the method further comprising:

receiving a query that specifies a relationship search that designates at least one of a term and an indication of a grammatical relationship associated with the designated term or a tag type that specifies a type of grammatical role, person, location, or thing;

translating the query to a set of Boolean expressions;

executing a search engine that evaluates the Boolean expressions against the enhanced data structures of the sentences to determine a set of sentence clauses that match the query; and

returning indications to the set of matching sentence clauses.

58. The memory medium of claim 57 wherein the received query specifies the relationship search by means of a natural language query that is transformed to the designated at least one of the term or the tag type.

59. The memory medium of claim 57 wherein the received query specifies a relationship search in combination with a document level Boolean search for at least one keyword to be located in each matching document.

60. The memory medium of claim 57 wherein the received query specifies a relationship search that is constrained by at least one of a meta-data tag expression or an expression that indicates a value of a prepositional phrase.

61. The memory medium of claim 57 wherein the tag type specifies at least one of an entity specification or a path in an ontology.

62. The memory medium of claim 57 wherein the relationship search specifies a search term using a wildcard.

63. The memory medium of claim 62 wherein the wildcard indicates a single character, range of characters, whole word, range of words, or a specific occurrence of a word.

64. The memory medium of claim 57 wherein the relationship search designates a value of at least one of a subject, an object, or a verb and the search engine determines all clauses in the corpus of documents where a grammatical relationship exists that satisfies the designated value of the at least one subject, object, or verb.

65. The memory medium of claim 64 wherein the relationship search designates a value of a subject and the search engine determines a corresponding object and a corresponding verb of all clauses that contain a subject having the designated value.

66. The memory medium of claim 64 wherein the relationship search designates a value of an object and the search engine determines a corresponding subject and a corresponding verb of all clauses that contain an object having the designated value.

67. The memory medium of claim 64 wherein the relationship search designates a value of a verb and the search engine determines a corresponding subject and a corresponding object of all clauses that contain a verb having the designated value or a similar verb to the designated value.

68. The memory medium of claim 64 wherein the relationship search designates a wildcard for at least one of the values of the designated at least one subject, object, or verb.

69. The memory medium of claim 64 wherein the search engine considers the presence in a sentence clause of a term used in a modifier grammatical role as a modifier of the designated value of the subject, object, or verb as a match to the designated value of the subject, object, or verb.

70. The memory medium of claim 57 wherein the search engine is an off-the-shelf keyword search engine.

71. The memory medium of claim 45, the method further comprising:

receiving a script that specifies a plurality of queries in a script language, each query specifying a relationship search that designates at least one of a term and an indication of a grammatical relationship that encompasses the term or a tag type that specifies a type of grammatical role, person, location, or thing;

translating the plurality of queries to a set of Boolean expressions;

executing a search engine that evaluates the Boolean expressions against the enhanced data structures of the sentences to determine a set of sentence clauses that match the Boolean expressions according to the script.

72. The memory medium of claim 71 wherein the script comprises at least one of control flow instructions, group constructs, query order, or functions.

73. The memory medium of claim 45 wherein the one or more inverted indexes of terms is a matrix that tracks occurrences of the terms across the corpus of documents.

74. The memory medium of claim 73 wherein the matrix is at least one of a term-document matrix, a term-sentence matrix, or a term-clause matrix.

75. The memory medium of claim 73 wherein the matrix is a plurality of term-clause matrices, each corresponding to a different grammatical role.

76. The memory medium of claim 73 wherein each indexed term and additional term is associated with a location that corresponds to a particular clause, sentence, and document.

77. The computer-readable memory medium of claim 45 wherein the memory medium is a memory in a computing system.

78. The computer-readable memory medium of claim 77 wherein the contents are instructions that, when executed, cause a computer processor in the computing device to perform the method.

79. A computing system that is configured to index a corpus of documents for electronic searching, each document having at least one sentence, each sentence having a plurality of terms, comprising:

a parser that is configured, when executed, to parse each sentence of each document to generate a dependency structure that specifies a plurality of syntactic elements that correspond to a plurality of the terms of the sentence and their grammatical relationship to each other;

a post processing module that is configured, when executed, to normalize the dependency structure to a plurality of tagged terms, each tagged term indicating an association between the term that corresponds to the syntactic element and an associated tag type, the associated tag type specifying a grammatical role of the corresponding term as used in the sentence, the grammatical role designating at least one of a subject, an object, a governing verb, a modifier, or a part of a prepositional phrase, wherein at least one of the tagged terms has an associated tag type that that specifies that the corresponding term is a subject or an object of the sentence, wherein at least one of the tagged terms has an associated tag type that specifies that the associated term of the sentence is a modifier of another term of the sentence that has an associated tag type that specifies that the another term is a subject, object, or verb of the sentence, and wherein at least one of the tagged terms has an associated tag type that additionally refers to an entity type that identifies the corresponding term as a type of person, place, or thing; and

a sentence transformation module that is configured, when executed, to transform the plurality of tagged terms to an enhanced data structure that stores and treats each tagged term as an encoded additional term of the sentence in one or more inverted indexes of terms annotated with relationship information, thereby enabling a search engine, to perform relationship searches by determining from the enhanced data structure whether a designated term having an associated tag type that specifies a desired grammatical role and/or a desired entity type is present in the sentence in a same role, in a manner similar to the manner the search engine uses to determine whether a designated term is present in the sentence, at least one of the relationship searches capable of returning a plurality of relationships between at least two entities as a result of a single search specification.

80. The system of claim 79 wherein the search engine is at least one of a keyword search or a Boolean search engine.

81. The system of claim 79 wherein the search engine performs string matching to determine whether the designated term having the associated tag type is present in the sentence in the same role.

82. The system of claim 79 wherein the search engine performs pattern matching to determine whether the designated term having the associated tag type is present in the sentence in the same role.

83. The system of claim 79 wherein the one or more inverted indexes of terms is an augmented term-document matrix configured to be searched using pattern matching or string matching to determine sentences that match a specified relationship query.

84. The system of claim 79 wherein the transforming each sentence to the enhanced data structure is performed for each clause of each sentence such that the tagged terms are treated as additional terms of each clause of the sentence and the search engine determines whether the designated syntactic term having the associated tag type is present in each clause.

85. The system of claim 79 wherein the at least one of the tagged terms that has the associated tag type that additionally refers to the entity type specifies an ontology path.

86. The system of claim 79 wherein the post processing module is configured, when executed, to normalize the dependency structure by applying linguistic normalization techniques to the plurality of syntactic elements to generate the plurality of tagged terms.

87. The system of claim 86 wherein the linguistic normalization techniques include applying at least one of a one transformational grammar rule, a coreference resolution rule, a verbalization rule, or a verb sense rule.

88. The system of claim 86 wherein the linguistic normalization techniques include applying at least one rule that normalizes a set of synonyms or acronyms to a standard term or phrase.

89. The system of claim 86 wherein the linguistic normalization techniques comprise identifying and generating tagged terms that include hypernyms and hyponyms.

90. The system of claim 79, further comprising:

a query interface module that is configured, when executed, to

receive a query that specifies a relationship search that designates at least one of a term and a grammatical relationship associated with the designated term or a tag type that specifies a type of grammatical role, person, place, or thing;

translate the query to a set of Boolean expressions;

execute a search engine that evaluates the Boolean expressions against the enhanced data structures of the sentences to determine a set of sentence clauses that match the query; and

return indications to the set of matching sentence clauses.

91. The system of claim 90 wherein the received query specifies the relationship search by means of a natural language query that is transformed to the designated at least one of the term or the tag type.

92. The system of claim 90 wherein the received query specifies a relationship search in combination with a document level Boolean search for at least one keyword to be located in each matching document.

93. The system of claim 90 wherein the received query specifies a relationship search that is constrained by at least one of a meta-data tag expression or an expression that indicates a value of a prepositional phrase.

94. The system of claim 90 wherein the relationship search specifies a search term using a wildcard.

95. The system of claim 90 wherein the relationship search designates a value of at least one of a subject, an object, or a verb and the search engine determines all clauses in the corpus of documents where a grammatical relationship exists that satisfies the designated value of the at least one subject, object, or verb.

96. The system of claim 95 wherein the search engine considers the presence in a sentence clause of a term used in a modifier grammatical role as a modifier of the designated value of the subject, object, or verb as a match to the designated value of the subject, object, or verb.

97. The system of claim 90 wherein the search engine is an off-the-shelf keyword search engine.

98. The system of claim 79, further comprising:

a query interface module that is configured, when executed, to

receive a script that specifies a plurality of queries in a script language, each query specifying a relationship search that designates at least one of a term and an indication of a grammatical relationship that encompasses the term or a tag type that specifies a type of grammatical role, person, place, or thing;

translate the plurality of queries to a set of Boolean expressions;

execute a search engine that evaluates the Boolean expressions against the enhanced data structures of the sentences to determine a set of sentence clauses that match the Boolean expressions according to the script.

99. The system of claim 98 wherein the script comprises at least one of control flow instructions, group constructs, query order, or functions.

100. The system of claim 79 wherein the one or more inverted indexes of terms a matrix that tracks occurrences of the terms across the corpus of documents.

101. The system of claim 100 wherein the matrix is at least one of a term-document matrix, a term-sentence matrix, or a term-clause matrix.

102. The system of claim 100 wherein the matrix is a plurality of term-clause matrices, each corresponding to a different grammatical role.

103. The system of claim 100 wherein each indexed term and additional term is associated with a location that corresponds to a particular clause, sentence, and document.

104. A computer-readable memory medium that contains a reverse index for storing a corpus of documents according to terms present in the documents, the index configured to be accessed by a computer processor that is controlled by search engine to match a relationship query against the corpus of documents using pattern or string matching, the index comprising:

a plurality of terms, each term of the plurality of terms indicating at least one sentence in which the term occurs; and

a plurality of tagged terms, each tagged term specifying a grammatical role that indicates a grammatical relationship of an associated term in the at least one sentence to other terms in the at least one sentence, each tagged term indicating the at least one sentence in which the associated term occurs, at least one of the tagged terms specifying a grammatical role that indicates that the associated term is a subject or an object, at least one of the tagged terms having an associated tag type that specifies that the associated term of the sentence is a modifier of another term of the sentence that has an associated tag type that specifies that the another term is a subject, object, or verb of the sentence, and at least one of the tagged terms additionally specifying a semantic tag that specifies that the associated term is a type of person, location, or thing;

such that the search engine can determine, by pattern matching query terms against the terms and tagged terms of the reverse index, a set of sentences that match a relationship indicated by the query.

105. The memory medium of claim 104 wherein the search engine is a keyword-style search engine.

106. The memory medium of claim 104 wherein the grammatical relationship indicated by each tagged term is at least one of a subject, an object, a governing verb, a modifier, or a part of a prepositional phrase.

107. The memory medium of claim 104 wherein the semantic tag is at least one of an entity tag or a path specification in a classification structure.

108. The memory medium of claim 104 wherein the reverse index is a term-clause index wherein each term indicates a clause within the indicated at least one sentence in which the term occurs.

109. The memory medium of claim 104 wherein the reverse index is a term-clause index wherein each tagged term indicates a clause within the indicated at least one sentence in which the associated term occurs.

110. A search engine configured, when executed, to process queries against a corpus of documents that are stored in the reverse index contained in the computer readable memory medium and structured according to claim 104.

111. The search engine of claim 110 wherein keyword searching techniques are performed to process queries.

112. A method in a computing system comprising storing a corpus of documents in the reverse index contained in the computer readable memory medium and structured according to claim 104.

113. The method of claim 112 wherein the reverse index is a term-clause index.

114. The method of claim 112 wherein the reverse index is a term-sentence index.

115. The method of claim 112 wherein the reverse index is a term-document index.