System for enhancing expert-based computerized analysis of a set of digital documents and methods useful in conjunction therewith

• US 9,881,080 B2
• Filed: 07/15/2016
• Issued: 01/30/2018
• Est. Priority Date: 04/22/2009
• Status: Active Grant

• Current Assignee: Microsoft Israel Research And Development (2002) Ltd (Microsoft Corporation)
• Original Assignee: Microsoft Israel Research And Development (2002) Ltd (Microsoft Corporation)
• Inventors: Ravid, Yiftach
• Primary Examiner(s): ALAM, SHAHID AL
• Application Number:
  US15/212,092
• Publication Number:
  US 20170011118A1
• Time in Patent Office: 564 Days
• Field of Search: 707740, 707749
• CPC Class Codes: G06F 16/3326 using relevance feedback from the user, e.g. relevance feedback on documents, documents sets, document terms or passages
• CPC Class Codes: G06F 16/35 Clustering; Classification
• CPC Class Codes: G06F 16/353 into predefined classes
• CPC Class Codes: G06F 16/93 Document management systems
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

An electronic document analysis method receiving N electronic documents pertaining to a case encompassing a set of issues including at least one issue and establishing relevance of at least the N documents to at least one individual issue in the set of issues, the method comprising, for at least one individual issue from among the set of issues, receiving an output of a categorization process applied to each document in training and control subsets of the at least N documents, the output including, for each document in the subsets, one of a relevant-to-the-individual issue indication and a non-relevant-to-the-individual issue indication; building a text classifier simulating the categorization process using the output for all documents in the training subset of documents; and running the text classifier on the at least N documents thereby to obtain a ranking of the extent of relevance of each of the at least N documents to the individual issue. The method may also comprise evaluating the text classifier'"'"'s quality using the output for all documents in the control subset.

Claims

1. A system comprising:

one or more processors; and

memory that stores instructions that are executable by the one or more processors to cause the system to perform operations comprising;

receiving a first output of a categorization process applied to a training subset of documents of a plurality of documents, the first output including a first indication and a second indication for each document in the training set of documents, the first indication indicating a relevance between a document in the training set of documents and an issue in a set of issues and the second indication indicating a lack of relevance between the document and the issue;

generating a classifier based at least in part on the first output;

executing the classifier on the plurality of documents to determine a second output, the second output indicating an extent of relevance of each document in the plurality of documents to the issue;

partitioning individual documents in the plurality of documents into subsets of documents based at least in part on the second output;

adding additional documents from at least one subset of the subsets of documents into the training subset of documents to generate a control subset of documents;

executing, as part of a first iteration, the classifier on the control subset of documents to determine a third output;

determining, based at least in part on the third output, a threshold associated with the classifier, the threshold being associated with a cutoff point for binarizing a ranking of individual documents in the control subset of documents;

computing, based at least in part on the cutoff point, a quality criterion associated with the classifier;

determining, based at least in part on the quality criterion, a first quality of performance of the classifier as applied to the control subset of documents;

receiving an input;

determining, based at least in part on the input, to initiate a second iteration;

determining a second quality of performance of the classifier for the second iteration; and

displaying a comparison of the first quality of performance and the second quality of performance.

2. The system as claim 1 recites, wherein the input comprises a user input received from a user.

3. The system as claim 1 recites, wherein the input comprises a computerized input.

4. The system as claim 1 recites, the operations further comprising:

generating at least one graph of respective quality criterions determined for the first iteration and the second iteration in view of respective iteration serial numbers for the first iteration and the second iteration; and

displaying the at least one graph to represent the comparison.

5. The system as claim 1 recites, the operations further comprising generating a computer display representative of the first output of the categorization process.

6. The system as claim 5 recites, wherein the computer display includes a histogram of ranks for each issue in the set of issues.

7. The system as claim 5 recites, wherein:

the computer display comprises a function of an indication of quality measures for at least the cutoff point associated with the first iteration and one or more other cutoff points associated with additional iterations.

8. The system as claim 7 recites, the operations further comprising:

determining at least one of an un-weighted F-measure, a weighted F-measure, a precision, a recall, or an accuracy associated with the first iteration; and

determining a quality measure of the quality measures associated with the cutoff point based on one or more of the un-weighted F-measure, the weighted F-measure, the precision, the recall, or the accuracy associated with the first iteration.

9. A method comprising:

receiving a first output of a categorization process applied to a training subset of documents of a plurality of documents, the first output including a first indication and a second indication for each document in the training set of documents, the first indication indicating a relevance between a document in the training set of documents and an issue in a set of issues and the second indication indicating a lack of relevance between the document and the issue;

generating a classifier based at least in part on the first output;

executing the classifier on the plurality of documents to determine a second output indicating an extent of relevance of each document in the plurality of documents to the issue;

partitioning, based at least in part on the second output, individual documents in the plurality of documents into subsets of documents;

adding additional documents from at least one subset of the subsets of documents into the training subset of documents to generate a control subset of documents;

executing, as part of a first iteration, the classifier on the control subset of documents to determine a third output;

determining, based at least in part on the third output, a threshold associated with the classifier, the threshold being associated with a cutoff point for binarizing a ranking of individual documents in the control subset of documents;

computing, based at least in part on the cutoff point, a quality criterion associated with the classifier;

determining, based at least in part on the quality criterion, a first quality of performance of the classifier as applied to the control subset of documents;

receiving an input;

determining, based at least in part on the input, to initiate a second iteration;

determining a second quality of performance of the classifier for the second iteration; and

displaying a comparison of the first quality of performance and the second quality of performance.

10. A computer storage device storing instructions that, when executed by one or more processors, cause a device to perform operations comprising:

receiving a first output of a categorization process applied to a training subset of documents of a plurality of documents, the first output including a first indication and a second indication for each document in the training set of documents, the first indication indicating a relevance between a document in the training set of documents and an issue in a set of issues and the second indication indicating a lack of relevance between the document and the issue;

generating a classifier based at least in part on the first output;

executing the classifier on the plurality of documents to determine a second output, the second output indicating an extent of relevance of each document in the plurality of documents to the issue;

partitioning individual documents in the plurality of documents into subsets of documents based at least in part on the second output;

adding additional documents from at least one subset of the subsets of documents into the training subset of documents to generate a control subset of documents;

executing, as part of a first iteration, the classifier on the control subset of documents to determine a third output;

determining, based at least in part on the third output, a threshold associated with the classifier, the threshold being associated with a cutoff point for binarizing a ranking of individual documents in the control subset of documents;

computing, based at least in part on the cutoff point, a quality criterion associated with the classifier;

determining, based at least in part on the quality criterion, a first quality of performance of the classifier as applied to the control subset of documents;

receiving an input;

determining, based at least in part on the input, to initiate a second iteration;

determining a second quality of performance of the classifier for the second iteration; and

displaying a comparison of the first quality of performance and the second quality of performance.

11. The computer storage device as claim 10 recites, wherein the input comprises a user input received from a user.

12. The computer storage device as claim 10 recites, wherein the input comprises a computerized input.

13. The computer storage device as claim 10 recites, the operations further comprising:

generating at least one graph of respective quality criterions determined for the first iteration and the second iteration in view of respective iteration serial numbers for the first iteration and the second iteration; and

displaying the at least one graph to represent the comparison.

14. The computer storage device as claim 10 recites, the operations further comprising generating a computer display representative of the first output of the categorization process.

15. The computer storage device as claim 14 recites, wherein:

the computer display comprises a function of an indication of quality measures for at least the cutoff point associated with the first iteration and one or more other cutoff points associated with additional iterations.