Error correction in tables using a question and answer system

• US 9,830,314 B2
• Filed: 11/18/2013
• Issued: 11/28/2017
• Est. Priority Date: 11/18/2013
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Byron, Donna K., Pikovsky, Alexander, Shivkumar, Abhishek, Winkler, Timothy P.
• Primary Examiner(s): Dorvil, Richemond
• Assistant Examiner(s): CHAVEZ, RODRIGO A
• Application Number:
  US14/082,862
• Publication Number:
  US 20150142418A1
• Time in Patent Office: 1,471 Days
• Field of Search: 704 7, 704 9, 704251
• CPC Class Codes: G06F 40/177 of tables; using ruled lines
• CPC Class Codes: G06F 40/232 Orthographic correction, e.g. spell checking or vowelisation
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06F 40/40 Processing or translation of natural language natural language analysis G06F40/20; semantic analysis G06F40/30

Abstract

Mechanisms are provided for performing tabular data correction in a document. The mechanisms receive a natural language document comprising a portion of content and analyze the portion of content within the natural language document to identify an erroneous sub-portion comprising an erroneous or missing item of information. The mechanisms generate a semantic signature for the erroneous sub-portion and generate a query based on the semantic signature. The mechanisms apply the query to a knowledge base to identify a candidate sub-portion of content. The mechanisms correct the erroneous sub-portion using the identified candidate sub-portion of content to generate a corrected natural language document.

Claims

1. A method, in a data processing system comprising a processor and a memory, for performing tabular data correction in a document, the method comprising:

configuring the processor to implement a natural language processing (NLP) system that performs natural language processing on natural language content at least by processing logical relationships in the natural language content;

responsive to receiving a natural language document with table structures and functional dependencies identified therein;

configuring an erroneous data value analysis engine to analyze a portion of content within a natural language document to identify an erroneous sub-portion within the natural language document comprising an erroneous or missing item of information;

analyzing, by the erroneous data value analysis engine executed by the processor in the data processing system, a portion of content within a natural language document to identify an erroneous sub-portion within the natural language document comprising an erroneous or missing item of information, wherein the portion of content comprises a table data structure present in the natural language document, wherein the erroneous sub-portion comprises a cell in the table data structure having an erroneous or missing data value, wherein the erroneous sub-portion is identified based on the erroneous sub-portion failing to conform with a regular structure associated with the portion of content within the natural language document, and wherein the regular structure is a repeatable pattern within the portion of content within the natural language document;

configuring a question generation engine to generate a semantic signature for the erroneous sub-portion;

generating, by the question generation engine executed by the processor in the data processing system, the semantic signature for the erroneous sub-portion, wherein generating the semantic signature for the erroneous sub-portion comprises;

performing, by the question generation engine, a discovery of functional dependencies between the erroneous or missing data value in the cell of the table data structure and a second portion of content, wherein the functional dependencies are indicated by the repeatable pattern;

analyzing, by the question generation engine, context information surrounding the erroneous sub-portion of content in the natural language document utilizing the functional dependencies between the erroneous or missing data value in the cell of the table data structure and a second portion of content; and

converting, by the question generation engine, the context information into a narrated statement using a narration mechanism;

configuring a Question and Answer (QA) system to generate a query based on the semantic signature and apply the query to a knowledge base to identify a candidate sub-portion of content for correcting the erroneous sub-portion;

generating, by the Question and Answer (QA) system executed by the processor in the data processing system, the query based on the semantic signature;

applying, by the QA system, the query to the knowledge base to identify the candidate sub-portion of content for correcting the erroneous sub-portion;

configuring a correction engine to correct the erroneous sub-portion to generate a corrected natural language document and store the corrected natural language document;

correcting, by the correction engine executed by the processor in the data processing system, the erroneous sub-portion using the identified candidate sub-portion of content to generate the corrected natural language document; and

storing, by the correction engine, the corrected natural language document in a storage device.

2. The method of claim 1, wherein the second portion of content comprises at least one of a cell in another table data structure of the natural language document, a cell in another table data structure of another natural language document, or a portion of content in another natural language document.

3. The method of claim 1, wherein performing discovery of functional dependencies between the erroneous or missing data value in the cell of the table data structure and the second portion of content comprises identifying one or more mathematical, statistical, logical, or conditional functions that operate on the cell of the table data structure and that indicate a functional dependency of the cell on the second portion of content.

4. The method claim 1, wherein performing discovery of functional dependencies between the erroneous or missing data value in the cell of the table data structure and the second portion of content comprises identifying functional relationships between headers of cells in the table data structure.

5. The method of claim 1, wherein correcting the erroneous sub-portion using the identified candidate sub-portion of content to generate a corrected natural language document comprises sending a notification of a user identifying the erroneous or missing item of information to the user and requesting that the user confirm replacement of the erroneous sub-portion of content with the candidate sub-portion of content.

6. The method of claim 1, wherein the knowledge base is a corpus of natural language documents, and wherein the query is applied to the corpus by the QA system to identify the candidate sub-portion of content.

7. The method of claim 6, wherein:

the QA system returns a plurality of candidate answers to the query,

identifying the candidate sub-portion of content comprises selecting, by the correction engine, a candidate answer from the plurality of candidate answers as the candidate sub-portion of content based on confidence measures for each of the plurality of candidate answers, and

correcting, by the correction engine, the sub-portion using the identified candidate sub-portion of content to generate a corrected natural language document comprises;

configuring the correction engine to compare the confidence measure for the selected candidate answer to a threshold confidence measure;

comparing, by the correction engine, the confidence measure for the selected candidate answer to the threshold confidence measure;

configuring the correction engine to perform one of automatic correction of the erroneous sub-portion using the identified candidate sub-portion or sending a notification to a user indicating the erroneous sub-portion, the candidate sub-portion, and requesting authorization to perform the correction, based on results of comparing the confidence measure for the selected candidate answer to the threshold confidence measure; and

performing, by the correction engine, one of the automatic correction of the erroneous sub-portion using the identified candidate sub-portion or sending the notification to the user indicating the erroneous sub-portion, the candidate sub-portion, and requesting authorization to perform the correction, based on the results of comparing the confidence measure for the selected candidate answer to the threshold confidence measure.

8. A computer program product comprising a non-transitory computer readable storage medium having a computer readable program for performing tabular data correction in a document stored therein, wherein the computer readable program, when executed on a computing device, causes the computing device to:

configure the computing device to implement a natural language processing (NLP) system that performs natural language processing on natural language content at least by processing logical relationships in the natural language content;

responsive to receiving a natural language document with table structures and functional dependencies identified therein;

configuring an erroneous data value analysis engine within the computing device to analyze a portion of content within a natural language document to identify an erroneous sub-portion within the natural language document comprising an erroneous or missing item of information;

analyze, by the erroneous data value analysis engine executed by a processor in the computing device, a portion of content within a natural language document to identify an erroneous sub-portion within the natural language document comprising an erroneous or missing item of information, wherein the portion of content comprises a table data structure present in the natural language document, wherein the erroneous sub-portion comprises a cell in the table data structure having an erroneous or missing data value, wherein the erroneous sub-portion is identified based on the erroneous sub-portion failing to conform with a regular structure associated with the portion of content within the natural language document, and wherein the regular structure is a repeatable pattern within the portion of content within the natural language document;

configure a question generation engine within the computing device to generate a semantic signature for the erroneous sub-portion;

generate, by the question generation engine executed by the processor in the computing device, the semantic signature for the erroneous sub-portion, wherein the computer readable program further causes the computing device to generate a semantic signature for the erroneous sub-portion at least by;

perform, by the question generation engine, a discovery of functional dependencies between the erroneous or missing data value in the cell of the table data structure and a second portion of content, wherein the functional dependencies are indicated by the repeatable pattern;

analyzing, by the question generation engine, context information surrounding the erroneous sub-portion of content in the natural language document; and

converting, by the question generation engine, the context information into a narrated statement using a narration mechanism;

configure a Question and Answer (OA) system within the computing device to generate a query based on the semantic signature and apply the query to a knowledge base to identify a candidate sub-portion of content for correcting the erroneous sub-portion;

generate, by the Question and Answer (QA) system executed by the processor in the computing device, the query based on the semantic signature;

apply, by the QA system, the query to the knowledge base to identify the candidate sub-portion of content for correcting the erroneous sub-portion;

configure a correction engine within the computing device to correct the erroneous sub-portion and store the corrected natural language document;

correct, by the correction engine executed by the processor in the computing device, the erroneous sub-portion using the identified candidate sub-portion of content to generate the corrected natural language document; and

store, by the correction engine, the corrected natural language document in a storage device.

9. The computer program product of claim 8, wherein the second portion of content comprises at least one of a cell in another table data structure of the natural language document, a cell in another table data structure of another natural language document, or a portion of content in another natural language document.

10. The computer program product claim 8, wherein the computer readable program further causes the computing device to perform discovery of functional dependencies between the erroneous or missing data value in the cell of the table data structure and the second portion of content at least by identifying functional relationships between headers of cells in the table data structure.

11. The computer program product of claim 8, wherein the computer readable program further causes the computing device to correct the erroneous sub-portion using the identified candidate sub-portion of content to generate a corrected natural language document at least by sending a notification of a user identifying the erroneous or missing item of information to the user and requesting that the user confirm replacement of the erroneous sub-portion of content with the candidate sub-portion of content.

12. The computer program product of claim 8, wherein the knowledge base is a corpus of natural language documents, and wherein the query is applied to the corpus by the QA system to identify the candidate sub-portion of content.

13. The computer program product of claim 12, wherein:

the QA system returns a plurality of candidate answers to the query,

identifying the candidate sub-portion of content comprises selecting, by the correction engine, a candidate answer from the plurality of candidate answers as the candidate sub-portion of content based on confidence measures for each of the plurality of candidate answers, and

correcting, by the correction engine, the sub-portion using the identified candidate sub-portion of content to generate a corrected natural language document comprises;

configuring the correction engine to compare the confidence measure for the selected candidate answer to a threshold confidence measure;

comparing, by the correction engine, the confidence measure for the selected candidate answer to the threshold confidence measure;

configuring the correction engine to perform one of automatic correction of the erroneous subportion using the identified candidate sub-portion or sending a notification to a user indicating the erroneous sub-portion, the candidate sub-portion, and requesting authorization to perform the correction, based on results of comparing the confidence measure for the selected candidate answer to the threshold confidence measure; and

performing, by the correction engine, one of the automatic correction of the erroneous sub-portion using the identified candidate sub-portion or sending the notification to the user indicating the erroneous sub-portion, the candidate sub-portion and requesting authorization to perform the correction, based on the results of comparing the confidence measure for the selected candidate answer to the threshold confidence measure.

14. An apparatus for performing tabular data correction in a document comprising:

a processor; and

a memory coupled to the processor, wherein the memory comprises instructions which, when executed by the processor, cause the processor to;

configure the processor to implement a natural language processing (NLP) system that performs natural language processing on natural language content at least by processing logical relationships in the natural language content;

responsive to receiving a natural language document with table structures and functional dependencies identified therein;

configuring an erroneous data value analysis engine to analyze a portion of content within a natural language document to identify an erroneous sub-portion within the natural language document comprising an erroneous or missing item of information;

analyze, by the erroneous data value analysis engine executed by a processor in the computing device, a portion of content within a natural language document to identify an erroneous sub-portion within the natural language document comprising an erroneous or missing item of information, wherein the portion of content comprises a table data structure present in the natural language document, wherein the erroneous sub-portion comprises a cell in the table data structure having an erroneous or missing data value, wherein the erroneous sub-portion is identified based on the erroneous sub-portion failing to conform with a regular structure associated with the portion of content within the natural language document, and wherein the regular structure is a repeatable pattern within the portion of content within the natural language document;

configure a question generation engine to generate a semantic signature for the erroneous sub-portion;

generate, by the question generation engine executed by the processor in the computing device, the semantic signature for the erroneous sub-portion, wherein the computer readable program further causes the computing device to generate a semantic signature for the erroneous sub-portion at least by;

perform, by the question generation engine, a discovery of functional dependencies between the erroneous or missing data value in the cell of the table data structure and a second portion of content, wherein the functional dependencies are indicated by the repeatable pattern;

analyzing, by the question generation engine, context information surrounding the erroneous sub-portion of content in the natural language document; and

converting, by the question generation engine, the context information into a narrated statement using a narration mechanism;

configure a Question and Answer (QA) system to generate a query based on the semantic signature and apply the query to a knowledge base to identify a candidate sub-portion of content for correcting the erroneous sub-portion;

generate, by the Question and Answer (QA) system executed by the processor in the computing device, the query based on the semantic signature;

apply, by the QA system, the query to the knowledge base to identify the candidate sub-portion of content for correcting the erroneous sub-portion;

configure a correction engine to correct the erroneous sub-portion and store the corrected natural language document;

correct, by the correction engine executed by the processor in the computing device, the erroneous sub-portion using the identified candidate sub-portion of content to generate the corrected natural language document; and

store, by the correction engine, the corrected natural language document in a storage device.