Combinatorial summarizer

Patent metadata

Field	Value
Publication number	US-9977829-B2
Application number	US-201214414479-A
Country	US
Kind code	B2
Filing date	Oct 12, 2012
Priority date	Oct 12, 2012
Publication date	May 22, 2018
Grant date	May 22, 2018

Abstract

A combinatorial summarizer includes a plurality of summarization engines, a processor in selective communication with each summarization engine, and computer readable instructions executable by the processor and embodied on a tangible, non-transitory computer readable medium. Each summarization engine is to select a respective plurality of sentences, and generate a relative rank and an associated weight for each sentence of the respective plurality of sentences. The computer readable instructions include instructions to determine a combined weight for each sentence of each respective plurality of sentences. The combined weight is based upon the respective associated weight and a respective relative human rank for each sentence in a set of sentences, including all sentences of each respective plurality of sentences. The computer readable instructions further include instructions to determine a total weight for each summarization engine based, respectively, upon the combined weights for each sentence of each respective plurality of sentences.

First claim

What is claimed is:

1. A combinatorial summarizer, comprising:
a plurality of summarization engines, each of which is a set of different computer readable instructions to:
select a respective plurality of sentences from a document; and
generate a relative rank and an associated weight for each sentence of the respective plurality of sentences;

a processor in selective communication with each of the plurality of summarization engines; and
computer readable instructions executable by the processor and embodied on a tangible, non-transitory computer readable medium, the computer readable instructions including:
computer readable instructions to determine a combined weight for each sentence of each of the respective plurality of sentences, the combined weight being based upon the respective associated weight and a respective relative human rank for each sentence in a set of sentences including all of the sentences of each of the respective plurality of sentences; and
computer readable instructions to determine a total weight for each of the plurality of summarization engines based, respectively, upon the combined weights for each sentence of each of the respective plurality of sentences.

2. The combinatorial summarizer as defined in claim 1 wherein the computer readable instructions further include computer readable instructions to apply a meta-algorithmic pattern to each sentence of each of the respective plurality of sentences.

3. The combinatorial summarizer as defined in claim 2 wherein the meta-algorithmic pattern is a voting pattern, and wherein the computer readable instructions to apply the voting pattern include:
computer readable instructions to determine a sum of the associated weights for each sentence of each of the respective plurality of sentences;
computer readable instructions to rank a predetermined number of a discrete subset of sentences of each of the respective plurality of sentences according to the sum of the associated weights;
computer readable instructions to assign a new weight to the sentences of the discrete subset; and
computer readable instructions to determine a total weight for a combination of the plurality of summarization engines based upon the new weight and the relative human rank for the sentences of the discrete subset.

4. The combinatorial summarizer as defined in claim 2 wherein the meta-algorithmic pattern is a weighted voting pattern, and wherein the computer readable instructions to apply the weighted voting pattern include:
computer readable instructions to determine a weight for each of the plurality of summarization engines;
computer readable instructions to determine a weighted sum for each sentence of each of the respective plurality of sentences, the weighted sum being based on the weights for each of the plurality of summarization engines and the associated weights for each sentence of each of the respective plurality of sentences;
computer readable instructions to rank a predetermined number of a discrete subset of sentences of each of the respective plurality of sentences according to the weighted sum;
computer readable instructions to assign a new weight to the sentences of the discrete subset; and
computer readable instructions to determine a total weight for a combination of the plurality of summarization engines based upon the new weight and the relative human rank for the sentences of the discrete subset.

5. The combinatorial summarizer as defined in claim 1 , further comprising:
an input device in selective communication with the processor, the input device to receive a human ranking for each sentence in the set of sentences; and
computer readable instructions for assigning the relative human rank to each sentence in the set of sentences based upon the human ranking.

6. The combinatorial summarizer as defined in claim 1 wherein the summarizer is part of a cloud computing system.

7. The combinatorial summarizer as defined in claim 1 wherein the computer readable instructions further include computer readable instructions to generate the set of sentences.

8. The combinatorial summarizer of claim 1 , wherein the relative rank from each of the plurality of summarization engines are different quantitative ranks and the human rank is a qualitative rank, and the rankings are from the most salient sentence to the least salient sentence, and wherein the human rank does not include ranking sentences of the document that were not ranked by any one of the summarization engines.

9. A combinatorial summarization method, comprising:
receiving, from each of a plurality of summarization engines, each of which is a set of different computer readable instructions and by a processor running computer readable instructions embodied on a tangible, non-transitory computer readable medium, a relative rank and an associated weight for each sentence of a respective plurality of sentences selected by each of the plurality of summarization engines from a document;
receiving, by the processor, a relative human rank for each sentence in a set of sentences including all of the sentences of each of the respective plurality of sentences;
determining, by the processor, a combined weight for each sentence of each of the respective plurality of sentences, the combined weight being based upon the respective associated weight and the respective relative human rank; and
determining, by the processor, a total weight for each of the plurality of summarization engines based, respectively, upon the combined weights for each sentence of each of the respective plurality of sentences.

10. The combinatorial summarization method as defined in claim 9 , further comprising evaluating the plurality of summarization engines based upon the total weights.

11. The combinatorial summarization method as defined in claim 9 , further comprising applying, by the processor, a meta-algorithmic pattern to each sentence of each of the respective plurality of sentences.

12. The combinatorial summarization method as defined in claim 11 , wherein the meta-algorithmic pattern is a voting pattern, and wherein applying the voting pattern includes:
determining, by the processor, a sum of the associated weights for each sentence of each of the respective plurality of sentences;
ranking, by the processor, a predetermined number of a discrete subset of sentences of each of the respective plurality of sentences according to the sum of the associated weights;
assigning, by the processor, a new weight to the sentences of the discrete subset; and
determining, by the processor, a total weight for a combination of the plurality of summarization engines based upon the new weight and the relative human rank for the sentences of the discrete subset.

13. The combinatorial summarization method as defined in claim 11 , wherein the meta-algorithmic pattern is a weighted voting pattern, and wherein applying the weighted voting pattern includes:
determining, by the processor, a weight for each of the plurality of summarization engines;
determining, by the processor, a weighted sum for each sentence of each of the respective plurality of sentences, the weighted sum being based on the weights for each of the plurality of summarization engines and the associated weights for each sentence of each of the respective plurality of sentences;
ranking, by the processor, a predetermined number of a discrete subset of sentences of each of the respective plurality of sentences according to the weighted sum;
assigning, by the processor, a new weight to the sentences of the discrete subset; and
determining, by the processor, a total weight for a combination of th

Assignees

• Hewlett Packard Development Co

Inventors

• Simske Steven J
• Sturgill Malgorzata M

Classifications

• G06F40/35

  Discourse or dialogue representation · CPC title

• G06F16/93

  Document management systems · CPC title

• G06F16/345 Primary

  Summarisation for human users · CPC title

• G06F16/3334

  Selection or weighting of terms from queries, including natural language queries · CPC title

• G06F16/3329

  Natural language query formulation · CPC title