Retrieving a specific sentence from a document

• US 9,767,186 B2
• Filed: 05/13/2014
• Issued: 09/19/2017
• Est. Priority Date: 06/11/2013
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Bhandari, Harendra, Nakakura, Kansaku, Wen, Lianzi
• Primary Examiner(s): Boccio, Vincent F
• Application Number:
  US14/275,994
• Publication Number:
  US 20140365475A1
• Time in Patent Office: 1,225 Days
• Field of Search: 707722, 707723, 707728-730, 707748, 707749, 707750
• CPC Class Codes: G06F 16/31 Indexing; Data structures therefor; Storage structures
• CPC Class Codes: G06F 16/3334 Selection or weighting of terms from queries, including natural language queries
• CPC Class Codes: G06F 16/338 Presentation of query results
• CPC Class Codes: G06F 16/35 Clustering; Classification

Abstract

An apparatus retrieves and outputs a specific sentence from a document. An acceptance hardware component acquires a retrieval key. An importance calculation hardware component calculates the level of importance of each sentence included in a search target document. A document processing hardware component extracts a sentence including the retrieval key from the search target. A retrieval result processing hardware component corrects a value of importance of each sentence calculated by the importance calculation hardware component for the sentence based on the position of the sentence in the search target document to calculate a corrected level of importance. A display control hardware component outputs the sentence extracted by the document processing hardware component according to the display order based on the corrected level of importance.

Claims

1. An apparatus for retrieving and outputting a specific sentence included in a document, the apparatus comprising:

a retrieval key acquiring hardware component for acquiring from a user an input of a retrieval key used to search a search target document, the retrieval key to determine a display sequence of sentences extracted from the search target document in response to the search;

an importance calculation hardware component for calculating an initial level of importance of each sentence included in the search target document, wherein the initial level of importance is determined by executing steps to;

determine, for a first sentence in the search target document, a row vector having elements corresponding to each sentence in the search target document, wherein a first element of the row vector has a first position in the row vector corresponding to a position of a second sentence in the search target document and a value corresponding to a relevance between the first sentence and the second sentence,

generate a matrix having rows corresponding to each sentence in the search target document, wherein a first row located at the first position amongst rows of the matrix is the first vector, and

generate an eigenvector corresponding to a maximum eigenvalue of the matrix, wherein an initial level of importance of the first sentence is a value of a second element of the eigenvector located at the first position in the eigenvector;

an extraction hardware component for searching the search target document using the retrieval key acquired by the retrieval key acquiring hardware component to extract a sentence including the retrieval key from the search target document in order to acquire position information on the extracted sentence in the document;

an importance correction hardware component for correcting a value of importance of each sentence calculated by the importance calculation hardware component for the sentence extracted by the extraction hardware component based on the position of the sentence in the search target document obtained by the extraction hardware component to calculate a corrected level of importance; and

a retrieval result outputting hardware component for outputting the sentence extracted by the extraction hardware component according to an order based on the corrected level of importance calculated by the importance correction hardware component.

2. The apparatus according to claim 1, wherein the relevance between the first sentence and the second sentence is determined based on based on a number of appearances of words included in the first sentence and in the second sentence across the document.

3. The apparatus according to claim 1, wherein the importance correction hardware component adds a larger weight value as each sentence is located more anteriorly in the search target document to correct the initial level of importance calculated for the sentence by the importance calculation hardware component.

4. The apparatus according to claim 1, wherein:

the retrieval key acquiring hardware component acquires from a user, as the retrieval key, a portion identified by specifying a part or a whole of one sentence included in the search target document, and

the importance correction hardware component adds a larger weight value as each sentence is located closer to a sentence having the portion identified as the retrieval key in the search target document to correct the initial level of importance calculated for the sentence by the importance calculation hardware component.

5. The apparatus according to claim 1, wherein the importance calculation hardware component calculates the relevance between the first sentence and the second sentence is determined based on based on a term frequency-inverse document frequency (tf-idf) of words included in the search target document.

6. An apparatus for retrieving and outputting a specific sentence included in a document, the apparatus comprising:

a display hardware component for displaying a document;

an acceptance hardware component for accepting a specification of a specific portion in the document displayed on the display hardware component;

an extraction hardware component for searching the document using, as a retrieval key, the specific portion in the document accepted by the acceptance hardware component to extract a sentence including the retrieval key from the search target document in order to acquire position information on the extracted sentence in the document;

a retrieval result processing hardware component for determining a display order of each sentence extracted by the extraction hardware component based on an initial level of importance determined for the sentence included in the document displayed on the display hardware component based on the number of appearances of a word included in the sentence across the document, and the position information on the sentence extracted by the extraction hardware component, wherein the initial level of importance is determined by executing steps to;

determine, for a first sentence in the document, a row vector having elements corresponding to each sentence in the document, wherein a first element of the row vector has a first position in the row vector corresponding to a position of a second sentence in the document and a value corresponding to a relevance between the first sentence and the second sentence,

generate a matrix having rows corresponding to each sentence in the document, wherein a first row located at the first position amongst rows of the matrix is the first vector, and

generate an eigenvector corresponding to a maximum eigenvalue of the matrix, wherein an initial level of importance of the first sentence is a value of a second element of the eigenvector located at the first position in the eigenvector; and

a display control hardware component for causing the display hardware component to display the sentence extracted by the extraction hardware component according to the display order determined by the retrieval result processing hardware component.

7. The apparatus according to claim 6, wherein the retrieval result processing hardware component determines the display order of each sentence extracted by the extraction hardware component based on a value obtained by adding a larger weight value to the initial level of importance of the sentence extracted by the extraction hardware component as the sentence is determined based on the position information to be located more anteriorly in the document.

8. The apparatus according to claim 6, wherein the retrieval result processing hardware component determines the display order of each sentence extracted by the extraction hardware component based on a value obtained by adding a larger weight value to the initial level of importance of the sentence extracted by the extraction hardware component as the sentence is determined based on the position information to be located closer in the document to a sentence with the retrieval key specified therein.

9. A method for retrieving a specific sentence from a document and causing a display hardware component to display the specific sentence on a computer including the display hardware component for displaying the document, the method comprising:

accepting, by one or more processors, an input of a retrieval key;

searching, by one or more processors, the document based on the accepted retrieval key to extract a sentence including the retrieval key from the search target document in order to acquire position information on the extracted sentence in the document, wherein the retrieval key determines a display sequence of sentences extracted from the search target document in response to the search;

determining, by one or more processors based on an initial level of importance determined based on the number of appearances of a word included in each sentence across the document and the position information on the sentence extracted based on the retrieval key, a display order of the sentence included in the document displayed on the display hardware component and extracted based on the retrieval key, wherein the initial level of importance is determined by executing steps to;

determine, for a first sentence in the document, a row vector having elements corresponding to each sentence in the document, wherein a first element of the row vector has a first position in the row vector corresponding to a position of a second sentence in the document and a value corresponding to a relevance between the first sentence and the second sentence,

generate a matrix having rows corresponding to each sentence in the document, wherein a first row located at the first position amongst rows of the matrix is the first vector, and

generate an eigenvector corresponding to a maximum eigenvalue of the matrix, wherein an initial level of importance of the first sentence is a value of a second element of the eigenvector located at the first position in the eigenvector; and

causing, by one or more processors, the display hardware component to display the sentence extracted based on the retrieval key according to the determined display order.