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

• US 20050267871A1
• Filed: 12/13/2004
• Published: 12/01/2005
• Est. Priority Date: 08/14/2001
• Status: Active Grant

• Current Assignee: Fiver LLC
• Original Assignee: Insightful Corporation (Cloud Software Group)
• Inventors: Liang, Jisheng, Brown, Matthew E., Dhillon, Navdeep S., Pochman, Lubos, Nguyen, Thien, Koperski, Krzysztof, Marchisio, Giovanni B., Tusk, Carsten
• Granted Patent: US 7,526,425 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• 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 to generate a parse structure having a plurality of syntactic elements that correspond to the terms of the sentence;

normalizing a plurality of the syntactic elements of the generated parse 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;

transforming each sentence to an enhanced data structure of terms, wherein the plurality of the tagged terms are treated as additional terms of the sentence, thereby enabling a search engine to determine from the enhanced data structure whether a designated term having an associated tag type is present in the sentence in a similar manner to the manner the search engine uses to determine whether a designated term is present in the sentence.

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.

4. The method of claim 1 wherein the enhanced data structure is an index structure that indexes the terms of the sentences along with the tagged terms of the sentence, thereby treating the terms and the tagged terms as searchable keywords.

5. The method of claim 1 wherein the enhanced data structure is an augmented term-document matrix.

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 associated tag type is at least one of an entity tag, an ontology path, a part-of-speech tag, a grammatical role tag, or an action attribute tag.

8. The method of claim 7 wherein the grammatical role tag is at least one of a subject, object, verb, noun phrase, or a modifier.

9. The method of claim 8 wherein the verb is a governing verb.

10. The method of claim 7 wherein the entity tag is part of a hierarchical organization of tags or is a specified ontology path.

11. The method of claim 1 wherein the normalizing the plurality of the syntactic elements of the generated parse structure comprises:

applying linguistic normalization techniques to a plurality of syntactic elements of the generated parse structure to generate 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.

12. The method of claim 11 wherein the linguistic normalization techniques include applying at least one of a transformational grammar rule, a morphological simplification rule, a coreference resolution rule, a verbalization rule, or a verb sense rule.

13. The method of claim 12 wherein the verbalization rule is at least one of a noun verbalization rule, an adjective verbalization rule, or an adverb verbalization rule.

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

15. The method of claim 12 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.

16. The method of claim 11 wherein the linguistic normalization techniques include applying at least one rule that normalize a set of synonyms and acronyms to a standard term or phrase.

17. The method of claim 11 wherein the linguistic normalization techniques comprise identifying and generating tagged terms that include hypernyms and hyponyms.

18. The method of claim 11 wherein the linguistic normalization techniques comprise identifying and generating tagged terms that include action attributes.

19. The method of claim 18 wherein the action attributes comprise a verb tense.

20. The method of claim 18 wherein the action attributes comprise a verb mood or modality indication such as whether the verb indicates a possibility, subjunctive, irrealis, negation, conditional, or causal relationship.

21. The method of claim 18 wherein the action attributes comprise similar verbs.

22. The method of claim 18 wherein the action attributes comprise troponyms, verb entailments, or hypernyms.

23. The method of claim 1, further comprising:

receiving a query that specifies a relationship search by designating at least one of a term, a tag type, or a tag value;

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.

24. The method of claim 23 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 a term, a tag type, or a tag value.

25. The method of claim 23 wherein the received query specifies a relationship search in combination with a document level Boolean search for at least one keyword.

26. The method of claim 23 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.

27. The method of claim 23 wherein the received query specifies a relationship search that is constrained by a meta-data tag expression.

28. The method of claim 23 wherein the received query specifies a relationship search that is constrained by an expression that indicates a value of a prepositional phrase.

29. The method of claim 23 wherein the tag type specifies at least one of a grammatical role, an entity specification, or a path in an ontology.

30. The method of claim 23 wherein the tag type specifies a subject, an object, or a verb.

31. The method of claim 23 wherein the tag value specifies a term that corresponds to at least one of a specified grammatical role or a path in an ontology.

32. The method of claim 23 wherein the relationship search specifies a search term using a wildcard.

33. The method of claim 32 wherein the wildcard indicates a single character, range of characters, whole word, range of words, or a specific occurrence of a word.

34. The method of claim 23 wherein the relationship search designates 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 at least one subject, object, or verb.

35. The method of claim 34 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.

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

37. The method of claim 34 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.

38. The method of claim 34 wherein the relationship search designates a wildcard for at least one of the values of the designated at least one subject, object, or verb.

39. The method of claim 34 wherein the search engine considers the presence in a sentence clause of a term having a modifier grammatical role that relates to the designated subject, object, or verb as a match to the designated subject, object, or verb.

40. The method of claim 23 wherein the search engine is an off- the-shelf keyword search engine.

41. 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, a tag type, or a tag value;

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.

42. The method of claim 41 wherein the script comprises at least one of control flow instructions, group constructs, query order, or functions.

43. The method of claim 1 wherein the terms of the sentence and the additional terms are indexed in a matrix that tracks occurrences of the terms across the corpus of documents.

44. The method of claim 43 wherein the matrix is at least one of a term-document matrix, a term-sentence matrix, or a term-clause matrix.

45. The method of claim 43 wherein the matrix is a plurality of term-clause matrices, each corresponding to a different grammatical role.

46. The method of claim 45 wherein the plurality of term-clause matrices comprise a subject index, an object index, and a verb index.

47. The method of claim 43 wherein each term with a designated tag type is associated with a location that corresponds to a particular clause, sentence, and document.

48. A computer-readable memory medium containing instructions that control a computer processor to index a corpus of documents for electronic searching, each document having at least one sentence, each sentence having a plurality of terms, by:

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;

normalizing a plurality of the syntactic elements of the generated parse 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;

transforming each sentence to an enhanced data structure of terms, wherein the plurality of the tagged terms are treated as additional terms of the sentence, thereby enabling a search engine, to determine from the enhanced data structure whether a designated term having an associated tag type is present in the sentence in a similar manner to the manner the search engine uses to determine whether a designated term is present in the sentence.

49. The memory medium of claim 48 wherein the search engine is at least one of a keyword search engine or a Boolean search engine.

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

51. The memory medium of claim 48 wherein the enhanced data structure is an index structure that indexes the terms of the sentences along with the tagged terms of the sentence, thereby treating the terms and the tagged terms as searchable keywords.

52. The memory medium of claim 48 wherein the enhanced data structure is an augmented term-document matrix.

53. The memory medium of claim 48 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.

54. The memory medium of claim 48 wherein the associated tag type is at least one of an entity tag, an ontology path, a part-of-speech tag, a grammatical role tag, or an action attribute tag.

55. The memory medium of claim 54 wherein the grammatical role tag is at least one of a subject, object, verb, noun phrase, or a modifier.

56. The memory medium of claim 54 wherein the entity tag is part of a hierarchical organization of tags or is a specified ontology path.

57. The memory medium of claim 48 wherein the normalizing the plurality of the syntactic elements of the generated parse structure comprises:

applying linguistic normalization techniques to a plurality of syntactic elements of the generated parse structure to generate 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.

58. The memory medium of claim 57 wherein the linguistic normalization techniques include applying at least one of a one transformational grammar rule, a morphological simplification rule, a coreference resolution rule, a verbalization rule, or a verb sense rule.

59. The memory medium of claim 57 wherein the linguistic normalization techniques include applying at least one rule that normalize a set of synonyms and acronyms to a standard term or phrase.

60. The memory medium of claim 57 wherein the linguistic normalization techniques comprise identifying and generating tagged terms that include hypernyms and hyponyms.

61. The memory medium of claim 48, further comprising instructions that control the computer processor by:

receiving a query that specifies a relationship search by designating at least one of a term, a tag type, or a tag value;

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.

62. The memory medium of claim 61 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 a term, a tag type, or a tag value.

63. The memory medium of claim 61 wherein the received query specifies a relationship search in combination with a document level Boolean search for at least one keyword.

64. The memory medium of claim 61 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.

65. The memory medium of claim 61 wherein the tag type specifies at least one of a grammatical role or an path in an ontology.

66. The memory medium of claim 61 wherein the tag value specifies a term that corresponds to at least one of a specified grammatical role, an entity specification, or a path in an ontology.

67. The memory medium of claim 61 wherein the relationship search specifies a search term using a wildcard.

68. The memory medium of claim 67 wherein the wildcard indicates a single character, range of characters, whole word, range of words, or a specific occurrence of a word.

69. The memory medium of claim 61 wherein the relationship search designates 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 at least one subject, object, or verb.

70. The memory medium of claim 69 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.

71. The memory medium of claim 69 wherein the relationship search designates a value of a object and the search engine determines a corresponding subject and a corresponding verb of all clauses that contain an object having the designated value.

72. The memory medium of claim 69 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.

73. The memory medium of claim 69 wherein the relationship search designates a wildcard for at least one of the values of the designated at least one subject, object, or verb.

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

75. The memory medium of claim 61 wherein the search engine is an off-the-shelf keyword search engine.

76. The memory medium of claim 48, further comprising instructions that control the computer processor by:

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, a tag type, or a tag value;

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.

77. The memory medium of claim 76 wherein the script comprises at least one of control flow instructions, group constructs, query order, or functions.

78. The memory medium of claim 48 wherein the terms of the sentence and the additional terms are indexed in a matrix that tracks occurrences of the terms across the corpus of documents.

79. The memory medium of claim 78 wherein the matrix is at least one of a term-document matrix, a term-sentence matrix, or a term-clause matrix.

80. The memory medium of claim 78 wherein the matrix is a plurality of term-clause matrices, each corresponding to a different grammatical role.

81. The memory medium of claim 78 wherein each term with a designated tag type is associated with a location that corresponds to a particular clause, sentence, and document.

82. A computer system that indexes 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 parsers each sentence of each document to generate a dependency structure that specifies a plurality of syntactic elements that correspond to the terms of the sentence and their relationship to each other;

a post processing module that is structured 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;

a sentence transformation module that is structured 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, thereby enabling a search engine, to determine from the enhanced data structure whether a designated term having an associated tag type is present in the sentence in a similar manner to the manner the search engine uses to determine whether a designated term is present in the sentence.

83. The system of claim 82 wherein the search engine is at least one of a keyword search or a Boolean search engine.

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

85. The system of claim 82 wherein the enhanced data structure is an index structure that indexes the terms of the sentences along with the tagged terms of the sentence, thereby treating the terms and the tagged terms as searchable keywords.

86. The system of claim 82 wherein the enhanced data structure is an augmented term-document matrix.

87. The system of claim 82 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.

88. The system of claim 82 wherein the associated tag type is at least one of an entity tag, an ontology path, a part-of-speech tag, a grammatical role tag, or an action attribute tag.

89. The system of claim 88 wherein the entity tag is part of a hierarchical organization of tags or is a specified ontology path.

90. The system of claim 82 wherein the post processing module is structured to normalize the plurality of the syntactic elements of the generated parse structure by applying linguistic normalization techniques to a plurality of syntactic elements of the generated parse structure to generate 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.

91. The system of claim 90 wherein the linguistic normalization techniques include applying at least one of a one transformational grammar rule, a morphological simplification rule, a coreference resolution rule, a verbalization rule, or a verb sense rule.

92. The system of claim 90 wherein the linguistic normalization techniques include applying at least one rule that normalize a set of synonyms and acronyms to a standard term or phrase.

93. The system of claim 90 wherein the linguistic normalization techniques comprise identifying and generating tagged terms that include hypernyms and hyponyms.

94. The system of claim 82, further comprising:

a query interface module that is structured to

receive a query that specifies a relationship search by indicating at least one of a term, a tag type, or a tag value;

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.

95. The system of claim 94 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 a term, a tag type, or a tag value.

96. The system of claim 94 wherein the received query specifies a relationship search in combination with a document level Boolean search for at least one keyword.

97. The system of claim 94 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.

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

99. The system of claim 94 wherein the relationship search designates 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 at least one subject, object, or verb.

100. The system of claim 99 wherein the search engine considers the presence in a sentence clause of a term having a modifier grammatical role that relates to the designated subject, object, or verb as a match to the designated subject, object, or verb.

101. The system of claim 94 wherein the search engine is an off- the-shelf keyword search engine.

102. The system of claim 82, further comprising:

a query interface module that is structured 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, a tag type, or a tag value;

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.

103. The system of claim 102 wherein the script comprises at least one of control flow instructions, group constructs, query order, or functions.

104. The system of claim 82 wherein the terms of the sentence and the additional terms are indexed in a matrix that tracks occurrences of the terms across the corpus of documents.

105. The system of claim 104 wherein the matrix is at least one of a term-document matrix, a term-sentence matrix, or a term-clause matrix.

106. The system of claim 104 wherein the matrix is a plurality of term-clause matrices, each corresponding to a different grammatical role.

107. The system of claim 104 wherein each term with a designated tag type is associated with a location that corresponds to a particular clause, sentence, and document.

108. A method in a computer system for performing a search of a corpus of documents, each document having at least one sentence, comprising:

receiving a search query that designates a desired grammatical relationship between a first entity and at least one of a second entity or an action;

transforming the search query into a Boolean expression;

determining a set of objects that match the Boolean expression using a keyword-style search of a data structure that indexes terms of the documents including grammatical relationship information; and

returning an indication of each matching object in the corpus that encompasses the desired relationship.

109. The method of claim 108 wherein the determining the set of object determines objects are at least one of clauses, sentences, paragraphs, or documents.

110. The method of claim 108 wherein the data structure stores the grammatical relationship information as additional terms of the documents.

111. The method of claim 108 wherein the designated at least one second entity or the action indicates a desire to match any second entity.

112. The method of claim 111, each sentence of each document comprising at least one clause, wherein the any second entity is any term used as a subject of a clause of a sentence.

113. The method of claim 111, each sentence of each document comprising at least one clause, wherein the any second entity is any term used as an object of a clause of a sentence.

114. The method of claim 111 wherein the first entity is any term that matches a specified entity type or ontology path specification.

115. The method of claim 108 wherein the first entity is any term that matches a specified entity type or ontology path specification.

116. The method of claim 108 wherein the designated at least one second entity or the action indicates a desire to match any action.

117. The method of claim 108 wherein the designated at least one second entity or the action is a verb.

118. The method of claim 117 wherein the returning the indication of each matching object that encompasses the desired relationship returns indications to objects that contain similar verbs to the designated verb.

119. The method of claim 117 wherein the returning the indication of each matching object that encompasses the desired relationship returns indications to objects that contain the same verb as the designated verb.

120. method of claim 117 wherein the returning the indication of each matching object that encompasses the desired relationship returns indications to objects that contain verbs of a similar classification to the designated verb.

121. The method of claim 108 wherein the designated at least one second entity or the action indicates a desire to match any action and a desire to match any second entity.

122. The method of claim 121 wherein the first entity is any term that matches a specified entity type.

123. The method of claim 108 wherein the receiving the search query that designates the desired grammatical relationship between a first entity and at least one of a second entity or an action specifies at least one of a prepositional constraint, a document keyword constraint, or a document metadata constraint.

124. The method of claim 108 wherein the search query includes a Boolean operation.

125. The method of claim 124 wherein the Boolean operation includes an AND, OR, or NOT operation.

126. The method of claim 108 wherein the search query includes an operator that specifies at least one of a proximity, a range, a wildcard, a weighted search based upon frequency, or a weighted keyword search operation.

127. The method of claim 108 wherein the search query includes a designation of at least one entity type.

128. The method of claim 127 wherein the at least one entity type is a path specification in a classification system.

129. The method of claim 127 wherein the at least one entity type is a path specification in a taxonomy that is specific to the corpus.

130. The method of claim 108 wherein the search query includes a wildcard specification in the designation of the desired grammatical relationship.

131. The method of claim 130 wherein the wildcard specification is one of a single character wildcard operator, a multi-character wildcard operator, or a word wildcard operator.

132. The method of claim 108, wherein the search query designates a desired grammatical relationship between the first entity and the second entity, the search query further designating a link entity specification that used to link the first entity and the second entity.

133. The method of claim 132 wherein the link entity specification is an entity type.

134. The method of claim 132 wherein the link entity specification is a path specification in a classification system.

135. The method of claim 108 wherein the transforming the search query to generate a Boolean expression incorporates transformational grammar rules to generate related grammatical relationships to search for.

136. The method of claim 108 wherein the generated Boolean expression includes an expression that causes a search for the desired grammatical relationship using at least one modifier.

137. The method of claim 136 wherein the at least one modifier is at least one of a subject modifier, an object modifier, a verb modifier, or an argument of preposition.

138. The method of claim 136 wherein the expression that causes a search for the desired grammatical relationship using the at least one modifier specifies an expression in which the modifier acts as a part of the first entity or the second entity.

139. The method of claim 136 wherein the expression that causes a search for the desired grammatical relationship using the at least one modifier specifies an expression in which the modifier acts as a part of the action.

140. The method of claim 108 wherein the data structure is a reverse index of terms that indexes at least one of documents, sentences, or clauses.

141. The method of claim 140 wherein the indexed terms include tagged terms that each denote a tag type associated with the term.

142. The method of claim 140 wherein the indexed terms include tagged terms that each denote a grammatical role associated with the term.

143. The method of claim 142 wherein the associated grammatical roles are at least one of subject, object, verb, or modifier.

144. The method of claim 140 wherein the indexed terms include tagged terms that each denote a semantic tag associated with the term.

145. The method of claim 144 wherein the associated semantic tags are path specifications in a classification system.

146. The method of claim 140 wherein the reverse index of terms comprises a plurality of reverse indices of terms.

147. The method of claim 108 wherein the data structure is at least one of a term-document matrix, a term-sentence matrix, or a term-clause matrix.

148. The method of claim 108 wherein the determining the set of sentences that match the Boolean expression performs pattern matching to determine the desired grammatical relationship.

149. The method of claim 108, the returning the indication of each matching object in the corpus that encompasses the desired relationship comprising:

returning an indication of at least one of each matching clause, each matching sentence, or each matching document in the corpus that encompasses the desired relationship.

150. The method of claim 108, the returning the indication of each matching object in the corpus that encompasses the desired relationship comprising:

in response to receiving a search query that designates a desired grammatical relationship between a first entity and any action, returning an indication of each matching object in the corpus that encompasses the first entity along with an indication of a corresponding action encompassed in the matching object.

151. The method of claim 108, the data structure that indexes terms of the documents including grammatical relationship information being stored across a plurality of storage repositories, wherein the determining the set of objects that match the Boolean expression using a keyword-style search of the data structure, further comprises:

performing a keyword-style search of the data structure against each storage repository that contains the portion of the index; and

merging the results of the search to return the indication of each matching object in the corpus that encompasses the desired relationship.

152. The method of claim 151 wherein the keyword-style searches against each storage repository that contains the portion of the index are performed using parallel processing techniques.

153. A computer-readable memory medium containing instructions to control a computer processor to search a corpus of documents, each document having at least one sentence, by:

receiving a search query that designates a desired grammatical relationship between a first entity and at least one of a second entity or an action;

transforming the search query into a Boolean expression;

determining a set of objects that match the Boolean expression using a keyword-style search of a data structure that indexes terms of the documents including grammatical relationship information; and

returning an indication of each matching object in the corpus that encompasses the desired relationship.

154. The memory medium of claim 153 wherein the determined objects are at least one of clauses, sentences, paragraphs, or documents.

155. The memory medium of claim 153 wherein the data structure stores the grammatical relationship information as additional terms of the documents.

156. The memory medium of claim 153 wherein the designated at least one second entity or the action indicates a desire to match any second entity.

157. The memory medium of claim 156, each sentence of each document comprising at least one clause, wherein the any second entity is any term used as a subject of a clause of a sentence.

158. The memory medium of claim 156, each sentence of each document comprising at least one clause, wherein the any second entity is any term used as an object of a clause of a sentence.

159. The memory medium of claim 156 wherein the first entity is any term that matches a specified entity type or ontology path specification.

160. The memory medium of claim 153 wherein the first entity is any term that matches a specified entity type or ontology path specification.

161. The memory medium of claim 153 wherein the designated at least one second entity or the action indicates a desire to match any action.

162. The memory medium of claim 153 wherein the designated at least one second entity or the action is a verb.

163. The memory medium of claim 162 wherein the returning the indication of each matching object that encompasses the desired relationship returns indications to objects that contain similar verbs to the designated verb.

164. The memory medium of claim 162 wherein the returning the indication of each matching object that encompasses the desired relationship returns indications to objects that contain the same verb as the designated verb.

165. memory medium of claim 162 wherein the returning the indication of each matching object that encompasses the desired relationship returns indications to objects that contain verbs of a similar classification to the designated verb.

166. The memory medium of claim 153 wherein the designated at least one second entity or the action indicates a desire to match any action and a desire to match any second entity.

167. The memory medium of claim 153 wherein the designated desired grammatical relationship specifies at least one of a prepositional constraint, a document keyword constraint, or a document metadata constraint.

168. The memory medium of claim 153 wherein the search query includes a Boolean operation.

169. The memory medium of claim 153 wherein the search query includes an operator that specifies at least one of a proximity, a range, a wildcard, a weighted search based upon frequency, or a weighted keyword search operation.

170. The memory medium of claim 153 wherein the search query includes a designation of at least one entity type or path specification in a classification system.

171. The memory medium of claim 153 wherein the search query includes a wildcard specification in the designation of the desired grammatical relationship.

172. The memory medium of claim 153, wherein the search query designates a desired grammatical relationship between the first entity and the second entity, the search query further designating a link entity specification that used to link the first entity and the second entity.

173. The memory medium of claim 153 wherein the transforming the search query to generate a Boolean expression incorporates transformational grammar rules to generate related grammatical relationships to search for.

174. The memory medium of claim 153 wherein the generated Boolean expression includes an expression that causes a search for the desired grammatical relationship using at least one modifier.

175. The memory medium of claim 153 wherein the data structure is a reverse index of terms that indexes at least one of documents, sentences, or clauses.

176. The memory medium of claim 175 wherein the indexed terms include tagged terms that each denote a tag type associated with the term.

177. The memory medium of claim 175 wherein the indexed terms include tagged terms that each denote a grammatical role associated with the term.

178. The memory medium of claim 175 wherein the indexed terms include tagged terms that each denote a semantic tag associated with the term.

179. The memory medium of claim 178 wherein the associated semantic tags are path specifications in a classification system.

180. The memory medium of claim 175 wherein the reverse index of terms comprises a plurality of reverse indices of terms.

181. The memory medium of claim 153 wherein the data structure is at least one of a term-document matrix, a term-sentence matrix, or a term-clause matrix.

182. The memory medium of claim 153 wherein the determining the set of sentences that match the Boolean expression performs pattern matching to determine the desired grammatical relationship.

183. The memory medium of claim 153, the returning the indication of each matching object in the corpus that encompasses the desired relationship comprising:

returning an indication of at least one of each matching clause, each matching sentence, or each matching document in the corpus that encompasses the desired relationship.

184. The memory medium of claim 153, the returning the indication of each matching object in the corpus that encompasses the desired relationship comprising:

in response to receiving a search query that designates a desired grammatical relationship between a first entity and any action, returning an indication of each matching object in the corpus that encompasses the first entity along with an indication of a corresponding action encompassed in the matching object.

185. The memory medium of claim 153, the data structure that indexes terms of the documents including grammatical relationship information being stored across a plurality of storage repositories, wherein the determining the set of objects that match the Boolean expression using a keyword-style search of the data structure, further comprises:

performing a keyword-style search of the data structure against each storage repository that contains the portion of the index; and

merging the results of the search to return the indication of each matching object in the corpus that encompasses the desired relationship.

186. The memory medium of claim 185 wherein the keyword-style searches against each storage repository that contains the portion of the index are performed using parallel processing techniques.

187. A search engine that searches a corpus of documents, each document having at least one sentence, comprising:

a data structure that indexes and stores terms of the documents along with annotations that include relationship information, each annotation associated with at least one term;

a keyword search engine that pattern matches an input string against the data structure and returns an indication of each matching object of the corpus; and

a query processor that is structured to

receive a relationship search query that is indicative of at least one syntactically or semantically annotated term;

transform the relationship search query into at least one Boolean expression;

invokes the keyword search engine to determine a set of objects that match the at least one Boolean expression by pattern matching the at least one annotated term indicated by the search query to the data structure, such that each matching object encompasses the relationship specified by the relationship search.

188. The search engine of claim 187 wherein the returned indications indicate at least one of clauses, sentences, paragraphs, or documents.

189. The search engine of claim 187 wherein the data structure stores the relationship information as additional terms of the documents.

190. The search engine of claim 187 wherein the relationship search query specifies a desired grammatical relationship between a first entity and at least one of a second entity or an action.

191. The search engine of claim 190 wherein the specified at least one second entity or the action indicates a desire to match any second entity.

192. The search engine of claim 190 wherein the first entity is any term that matches a specified entity type or ontology path specification.

193. The search engine of claim 190 wherein the specified at least one second entity or the action indicates a desire to match any action.

194. The search engine of claim 190 wherein the specified at least one second entity or the action is a verb.

195. The search engine of claim 190 wherein the specified at least one second entity or the action indicates a desire to match any action and a desire to match any second entity.

196. The search engine of claim 187, the relationship search specifying a desired action, wherein the returned indications of each matching object that encompasses the relationship returns indications to objects that contain similar verbs to a verb indicated by the desired action, the same verb as the verb indicated by the desired action, or a verb of a classification related to the verb indicated by the desired action.

197. The search engine of claim 187 wherein the relationship search query specifies at least one of a prepositional constraint, a document keyword constraint, or a document metadata constraint.

198. The search engine of claim 187 wherein the relationship search query includes a Boolean operation.

199. The search engine of claim 187 wherein the relationship search query includes an operator that specifies at least one of a proximity, a range, a wildcard, a weighted search based upon frequency, or a weighted keyword search operation.

200. The search engine of claim 187 wherein the relationship search query specifies at least one entity type or path specification in a classification system.

201. The search engine of claim 187 wherein the relationship search query includes a wildcard specification.

202. The search engine of claim 187, wherein the relationship search query includes a link entity specification.

203. The search engine of claim 187 wherein the transformed relationship search query incorporates transformational grammar rules.

204. The search engine of claim 187 wherein the transformed relationship search query includes an expression that causes a search using at least one modifier.

205. The search engine of claim 187 wherein the annotations that include relationship information denote a grammatical role of each associated term.

206. The search engine of claim 187 wherein the annotations denote semantic tags associated with the terms.

207. The search engine of claim 206 wherein the associated semantic tags are path specifications in a classification system.

208. The search engine of claim 187 wherein the data structure is a reverse index of terms that indexes at least one of documents, sentences, or clauses.

209. The search engine of claim 208 wherein the reverse index of terms comprises a plurality of reverse indices of terms.

210. The search engine of claim 187 wherein the data structure is at least one of a term-document matrix, a term-sentence matrix, or a term-clause matrix.

211. The search engine of claim 187, wherein the returned indication of each matching object returns an indication of at least one of each matching clause, each matching sentence, or each matching document in the corpus that encompasses the desired relationship.

212. The search engine of claim 187, the data structure that indexes and stores terms of the documents storing and indexing the terms with the annotations across a plurality of storage repositories, and wherein the keyword search engine performs pattern match searches of the input string against each storage repository that contains the portion of the index and merges the results of the pattern match searches to return the indication of each matching object in the corpus that encompasses the desired relationship.

213. The search engine of claim 212 wherein the pattern match searches of the input string against each storage repository that contains the portion of the index are performed using parallel processing techniques.

214. A computer-readable memory medium containing structured data that stores a syntactic query, the query executed by a computer processor under the control of a search engine to search a corpus of objects for objects that match the query, comprising:

a base component that specifies values for desired relationship parameters;

a prepositional constraint component that specifies a desired value for a prepositional phrase;

a keyword constraint component that specifies desired keyword values; and

a metadata constraint component that specifies desired values of metadata associated with each matching object,

whereby, when the search engine causes the search to be executed, objects that match the constraints specified by the base component, the prepositional constraint component, the keyword constraint component, and the metadata constraint component are determined to satisfy the query.

215. The memory medium of claim 214 wherein one or more of the components of the syntactic query are optional.

216. The memory medium of claim 214 wherein at least one of the components of the syntactic query is specified.

217. The memory medium of claim 214 wherein at least one of the components of the syntactic query contains a Boolean expression.

218. The memory medium of claim 214 wherein the base component specifies the desired relationship parameters in a general syntactic form:

Entity1 Directional-operator1 Action Directional-operator2 Entity2

wherein at least one of Entity1, Entity2, and Action parameters contains a non null value that indicates a search term, the Directional-operator1 parameter specifies the direction of the relationship between the Entity1 and the Action parameters, and the Directional-operator2 parameter specifies the direction of the relationship between the Entity2 and the Action parameters.

219. The memory medium of claim 218 wherein a value of the Directional-operator parameter is one of a greater-than symbol (“

>

”

), a right arrow symbol (“

→

”

), a less-than symbol (“

<

”

),a left arrow symbol (“

←

”

) or a combination indicating a bi-directional relationship (“

<

>

”

or “

⇄

”

).

220. The memory medium of claim 218 wherein a specification of a value of “

>

”

or “

→

”

for the Directional-operator1 parameter indicates that the value indicated by the Entity1 parameter is a subject of the value indicated by the Action parameter.

221. The memory medium of claim 218 wherein a specification of a value of “

<

”

or “

←

”

for Directional-operator1 parameter indicates that the value indicated by the Entity1 parameter is an object of the value indicated by the Action parameter.

222. The memory medium of claim 218 wherein a specification of a value of “

>

”

or “

→

”

for Directional-operator2 parameter indicates that the value indicated by the Entity2 parameter is an object of the value indicated by the Action parameter.

223. The memory medium of claim 218 wherein a specification of a value of “

<

”

or “

←

”

for Directional-operator2 parameter indicates that that the value indicated by the Entity2 parameter is a subject of the value indicated by the Action parameter.

224. The memory medium of claim 218 wherein a value for the Action parameter indicates a search term that represents at least one of a particular verb, similar verbs, or an action type.

225. The memory medium of claim 218 wherein a value for the Action parameter that is in the form of a quoted verb indicates a particular verb;

a value for the Action parameter that in the form of an unquoted verb indicates similar verbs to that which is specified; and

a value for the Action parameter that is in the form of a bracketed verb indicates an action type.

226. The memory medium of claim 218 wherein a value for the Entity1 or the Entity2 parameter is a noun or noun phrase.

227. The memory medium of claim 218 wherein a value for the Entity1 or the Entity2 parameter is a modifier.

228. The memory medium of claim 214 wherein the prepositional constraint component comprises the phrase “

PREP CONTAINS”

or the character “

{circumflex over ( 

)}”

followed by at least one search term.

229. The memory medium of claim 214 wherein the keyword constraint component comprises the phrase “

DOCUMENT CONTAINS”

or the character “

;

”

followed by at least one search term.

230. The memory medium of claim 214 wherein the metadata constraint component comprises the phrase “

METADATA CONTAINS”

or the character “

#”

followed by at least one expression that specifies a desired value for a metadata variable.

231. The memory medium of claim 214 wherein a wildcard can be specified as the value of a search term or a parameter of the base component.

232. The memory medium of claim 231 wherein the wildcard is at least one of the characters “

*”

or “

?”

.

233. The memory medium of claim 214 wherein curly braces are used to indicate indirect link searches.

234. The memory medium of claim 214 wherein square brackets are used to indicate an action type or an entity classification.

235. A search engine programmed to process syntactic queries that are stored in the structured data contained in the computer readable memory medium and structured according to claim 214.

236. A method in a computer system comprising storing syntactic queries in the structured data contained in the computer readable memory medium and structured according to claim 214.

246. A method in a computer system comprising storing a corpus of documents in the reverse index contained in the computer readable memory medium and structured according to claim 214.

247. The method of claim 246 wherein the reverse index is a term-clause index.

248. The method of claim 246 wherein the reverse index is a term-sentence index.

249. The method of claim 246 wherein the reverse index is a term-document index.

237. 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 accessed by a computer processor that is controlled by search engine to match a query against the corpus of documents, the index comprising:

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

a plurality of tagged terms, each tagged term specifying a syntactic role that is associated with the term in the at least one sentence and each tagged term indicating the at least one sentence in which the associated term occurs;

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

238. The memory medium of claim 237 wherein the search engine is a keyword-style search engine.

239. The memory medium of claim 237 wherein the syntactic role specified by each tagged term is at least one of a subject, object, verb, or modifier.

240. The memory medium of claim 237 wherein a plurality of the tagged terms specify a semantic tag that is associated with the term in the at least one sentence.

241. The memory medium of claim 240 wherein the semantic tag is at least one of an entity tag or a path specification in a classification structure.

242. The memory medium of claim 237 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.

243. The memory medium of claim 237 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.

244. A search engine programmed 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 237.

245. The search engine of claim 244 wherein keyword searching techniques are performed to process queries.