# US 2019/0095522 A1 · Search indexing using discourse trees

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

- Number: US 2019/0095522 A1
- Document ID: US20190095522A1
- Status: Published application
- Kind: A1
- Application number: 16/142,759
- Filed: 2018-09-26
- Published: 2019-03-28
- Earliest claimed: 2017-09-27
- Assignees: Oracle International Corporation
- Applicants: Oracle International Corporation
- Inventors: Boris Galitsky
- Web page: https://patentgrep.com/patent/US20190095522A1

## CPC

- G06F 16/322: {Trees}
- G06F 16/313: {Selection or weighting of terms for indexing}
- G06F 16/3329: Natural language query formulation
- G06F 16/3331: Query processing
- G06F 40/211: Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
- G06F 40/35: Discourse or dialogue representation
- G06F 40/30: Semantic analysis
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Systems, devices, and methods of the present invention create a searchable index that includes informative portions of text. In an example, a computer-implemented method creates a discourse tree from a body of text. For each non-terminal node in the discourse tree, the method identifies a rhetorical relationship associated with the non-terminal node. The method labels each terminal node associated with the non-terminal node as either a nucleus or a satellite. The method further accesses a rule associated with the rhetorical relationship, and selects, based on the rule, selects the fragment associated with the nucleus. The method creates a searchable index including the selected fragments.

## Claims

### Claim 1

A method of indexing text, the method comprising:
accessing a body of text comprising fragments;
creating a discourse tree from the body of text, wherein the discourse tree comprises a plurality of nodes, each non-terminal node representing a rhetorical relationship between two of the fragments, and each terminal node is associated with one or more fragments and is associated with a non-terminal node;
for each non-terminal node:
identifying a rhetorical relationship associated with the non-terminal node,
labeling each terminal node associated with the non-terminal node as either a nucleus or a satellite,
accessing a rule associated with the rhetorical relationship, wherein the rule specifies accessing a fragment of text associated with one or more of (i) the nucleus or (ii) the satellite, and
selecting, based on the rule, one or more of (i) the fragment of text associated with the nucleus or the (ii) fragment of text associated with the satellite; and
creating a searchable index comprising multiple entries, each entry corresponding to a selected fragment.

### Claim 2 (depends on 1)

The method of claim 1, further comprising:
determining, for each fragment, an additional question;
receiving a query from an external device;
searching, in each of the additional questions for matches corresponding to the query; and
responsive to determining a match in the additional questions, providing a response to the external device.

### Claim 3 (depends on 1)

The method of claim 1, wherein each entry represents a question that can be answered by the body of text or a phrase to be matched to the question.

### Claim 4 (depends on 1)

The method of claim 1, wherein the rule specifies (i) selecting the fragments associated with a nucleus and (ii) discarding fragments associated with a satellite.

### Claim 5 (depends on 1)

The method of claim 1, wherein one or more fragments comprise clauses of compound sentences.

### Claim 6 (depends on 1)

The method of claim 1, wherein each rhetorical relationship is one of an elaboration, enablement, condition, contrast, or attribution.

### Claim 7 (depends on 1)

The method of claim 1, further comprising:
determining, for each fragment, an additional question; and
providing the additional questions to an external device.

### Claim 8 (depends on 1)

The method of claim 1, further comprising, for at least one non-terminal node:
identifying the rhetorical relationship as either (i) a same unit or a (ii) joint relationship; and
responsive to determining that the rule specifies accessing both (i) the nucleus and (ii) the satellite, selecting both (i) the nucleus and (ii) the satellite.

### Claim 9 (depends on 1)

The method of claim 1, further comprising, for at least one non-terminal node:
identifying the rhetorical relationship as an attribution relationship, and
responsive to determining that the rhetorical relationship does not represent a query by an author, selecting the nucleus.

### Claim 10

A method comprising:
accessing a question-answer index, wherein each answer comprises fragments;
creating, for each answer, a discourse tree, wherein the discourse tree comprises a plurality of nodes, each non-terminal node representing a rhetorical relationship between two of the fragments and each terminal node is associated with one or more fragments;
for each non-terminal node in each answer:
identifying a rhetorical relationship associated with the non-terminal node,
labeling each terminal node associated with the non-terminal node as either a nucleus or a satellite,
accessing a rule associated with the rhetorical relationship, wherein the rule specifies accessing a fragment of text associated with either the nucleus or the satellite, and
selecting, based on the rule, one or more of (i) the fragment of text associated with the nucleus or (ii) the fragment of text associated with the satellite; and
creating an index of additional questions comprising multiple entries, each entry corresponding to an answer found in one of the selected fragments.

### Claim 11 (depends on 10)

The method of claim 10, method further comprising:
receiving a query from an external device;
searching, in each question of the question-answer index, for a match corresponding to the query;
responsive to determining that fewer than a threshold number of matches are available, searching for additional matches to the query in the index of additional questions;
responsive to determining that fewer than an additional threshold number of additional matches are available in the index of additional questions, searching for further matches in each answer in the question-answer index; and
providing the matches to the external device.

### Claim 12 (depends on 10)

The method of claim 10, further comprising, for at least one non-terminal node:
identifying the rhetorical relationship as either (i) a same unit or a (ii) joint relationship, and
selecting both (i) the nucleus and (ii) the satellite, wherein the rule specifies accessing both (i) the nucleus and (ii) the satellite.

### Claim 13 (depends on 10)

The method of claim 10, further comprising, for at least one non-terminal node:
identifying the rhetorical relationship as an attribution relationship, and
responsive to determining that the rhetorical relationship does not represent a query by an author, selecting the nucleus.

### Claim 14 (depends on 10)

The method of claim 10, wherein each entry represents an additional question that can be answered by the respective answer.

### Claim 15 (depends on 10)

The method of claim 10, wherein the rule specifies (i) selecting the fragments associated with a nucleus and (ii) discarding fragments associated with a satellite.

### Claim 16 (depends on 10)

The method of claim 10, wherein one or more fragments comprise clauses of compound sentences.

### Claim 17 (depends on 10)

The method of claim 10, wherein each rhetorical relationship is one of an elaboration, enablement, condition, contrast, or attribution.

### Claim 18

A method of training a classification model, the method comprising:
accessing a set of training data comprising a set of training data pairs, wherein each training data pair comprises text and an expected classification and wherein the set of training data includes both (i) a first training data pair that has an expected classification of a nucleus and (ii) a second training data pair that has an expected classification of a satellite; and
training a classification model by iteratively:
providing one of the training data pairs to the classification model;
receiving, from the classification model, a determined classification;
calculating a loss function by calculating a difference between the determined classification and the expected classification; and
adjusting internal parameters of the classification model to minimize the loss function.

### Claim 19 (depends on 18)

The method of claim 18, further comprising:
accessing a body of text comprising fragments;
creating a discourse tree from the body of text, wherein the discourse tree comprises a plurality of nodes, each non-terminal node representing a rhetorical relationship between two of the fragments and each terminal node associated with one or more fragments;
for each non-terminal node:
identifying a rhetorical relationship associated with the non-terminal node;
providing a fragment of text associated with the non-terminal node and the rhetorical relation to the classification model; and
receiving, from the classification model, a classification indicating whether the fragment of text is informative or not informative; and
creating a searchable index of additional questions comprising each of the fragments identified as informative.

### Claim 20 (depends on 19)

The method of claim 19, further comprising:
receiving a query from an external device;
searching, in each of question of a plurality of question-answer pairs, for matches corresponding to the query;
responsive to determining that fewer than a threshold number of matches are available, searching, in the searchable index of additional questions, for additional matches to the query;
responsive to determining that fewer than a threshold number of additional matches are available in the searchable index of additional questions, searching for further matches in each answer of the plurality of question-answer pairs; and
providing a response to the external device.
