# US 6,732,097 B1 · Relational text index creation and searching

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 6,732,097 B1
- Document ID: US6732097B1
- Status: Granted
- Kind: B1
- Application number: 09/927,580
- Filed: 2001-08-10
- Granted: 2004-05-04
- Earliest claimed: 2000-08-11
- Assignees: Attensity Corporation
- Inventors: Todd D. Wakefield; David L. Bean; Jeffrey A. Lorenzen
- Web page: https://patentgrep.com/patent/US6732097B1

## CPC

- G06F 16/316: {Indexing structures}
- Y10S 707/99934: Query formulation, input preparation, or translation
- Y10S 707/99935: Query augmenting and refining, e.g. inexact access
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- Y10S: TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS

## Abstract

In an environment where it is desire to perform information extraction over a large quantity of textual data, methods, tools and structures are provided for building a relational text index from the textual data and performing searches using the relational text index.

## Claims

### Claim 1

A method for creating a relational text index, searching system the method comprising the steps of:
accessing a group of documents containing natural language text,
parsing said document to identify grammatical parts of sentences in said documents,
applying caseframes to said parsed sentences to generate caseframe extractions, caseframes being syntactic structures that recognize local area context,
performing thematic role assignment on said caseframe extractions to generate thematic role extractions,
performing unification for each sentence that generates more than one thematic role extraction to generate a single unified representation of each sentence,
utilizing sentence information to build a relational text index that is usable by a computer system,
performing a search of said documents using said relational text index, and;
using a thematic role specifieras a search determiner.

### Claim 2 (depends on 1)

A method as recited in claim 1 further comprising the step of creating meta types for thematic roles.

### Claim 3 (depends on 1)

A method as recited in claim 1 further comprising using meta types as search determiners.

### Claim 4 (depends on 1)

A method as recited in claim 1 wherein said search step includes searching both based on a thematic role and on its root form.

### Claim 5 (depends on 1)

A method as recited in claim 1 further comprising permitting a user to input a search query said search query including a thematic role selected from the group consisting of actor, action, object and specifier.

### Claim 6 (depends on 1)

A method as recited in claim 1 wherein said search step includes comparing an actor to noun phrases in said relational text index.

### Claim 7 (depends on 1)

A method as recited in claim 1 wherein said search step includes comparing an action to verb phrases in said relational text index.

### Claim 8 (depends on 1)

A method as recited in claim 1 wherein said search step includes comparing said specifier to noun phrase specifiers in said relational text index.

### Claim 9 (depends on 1)

A method as recited in claim 1 further comprising accepting a search query from a user, said search query including terms in a mode selected from the group consisting of union mode and intersection mode.

### Claim 10

A method for creating a relational text index, searching system the method comprising the steps of:
accessing a group of documents containing a natural language text,
parsing said document to identify grammatical parts of sentences in said documents,
applying caseframes to said parsed sentences to generate caseframe extractions, caseframes being syntactic structures that recognize local area context,
performing thematic role assignment on said caseframe extractions to generate thematic role extractions,
performing unification for each sentence that generates more than one thematic role extraction to generate a single unified representation of each sentence,
utilizing sentence information to build a relational text index that is usable by a computer system,
accepting a search query from a user, said search query including actor and action roles,
performing a search of said documents using said relational text index by
querying said relational text index for instances when said actor or action is recorded in its appropriate role,
for each query match, retrieve the noun phrase of the extracted term,
for each query match, retrieve the document and phrase where the match occurred, and display search results to the user.

### Claim 11 (depends on 1)

A method as recited in claim 1 further comprising placing the relational text index into a database management system for later use.

### Claim 12 (depends on 1)

A method as recited in claim 1 further comprising:
accepting a search query from a user, and
collapsing on the root form of search terms.

### Claim 13 (depends on 1)

A method as recited in claim 1, further comprising the step of comparing verb roles to pre-defined meta types.

### Claim 14

A method for creating a relational text index search system, the method comprising the steps of:
accessing a corpus of natural language text documents,
for a plurality of said documents, parsing sentences in said documents to generate diagrammed sentences, said parsing step producing an output selected from the group consisting of noun phrases, verb phrases, prepositional phrases, adverbial phrases, adjectival phrases, clauses, and combinations of them,
applying caseframes to said diagrammed sentences to generate caseframe extractions,
performing thematic role assignment on said caseframe extractions to generate thematic role extractions, said thematic role assignment being performed by translating raw caseframe-extractions to specific thematic roles, and accessing a relational text index file,
appending thematic role information to said relational text index file,
performing a search of said corpus of natural language documents using said relational text index,
accepting a search query from a user and permitting a user to specify at least one search role selected from the group consisting of actor, action, object and specifier, and
comparing a user's search roles to thematic roles in said relational text index.

### Claim 15 (depends on 14)

A method as recited in claim 14 further comprising collapsing on morphological root forms of search roles.

### Claim 16 (depends on 14)

A method as recited in claim 14 further comprising displaying portions of documents where user search roles were found.
