Specialized language identification

• US 10,216,721 B2
• Filed: 09/30/2014
• Issued: 02/26/2019
• Est. Priority Date: 09/30/2014
• Status: Active Grant

• Current Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Original Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Inventors: Simske, Steven J, Riss, Marcelo, Vans, Marie, Sturgill, Malgorzata M
• Primary Examiner(s): Azad, Abul K
• Application Number:
  US15/515,933
• Publication Number:
  US 20170300471A1
• Time in Patent Office: 1,610 Days
• Field of Search: None
• CPC Class Codes: G06F 16/3334 Selection or weighting of terms from queries, including natural language queries
• CPC Class Codes: G06F 16/3344 using natural language analysis
• CPC Class Codes: G06F 40/263 Language identification
• CPC Class Codes: G06F 40/279 Recognition of textual entities

Abstract

Examples herein disclose multiple engines to produce output representative of a summary of document produced by each of the multiple engines. The examples apply a weighting mechanism to the output specific to that engine to obtain a value corresponding to that output. The examples identify specialized language if the value corresponding to that output reaches at least a particular threshold.

Claims

1. A system comprising:

multiple engines that are each to produce output representative of a summary of the document, wherein each one of the multiple engines applies a different type of engine selected from a group of engines comprising an extractive type of engine, an abstractive type of engine, and a frequency type of engine, wherein the output from each of the multiple engines varies between the multiple engines in accordance with a respective type of engine;

a composite engine to generate a filtered set of content in a single output to reduce a size of the output produced by the multiple engines, wherein the filtered set of content comprises different combinations of the output from the multiple engines that have different densities of specialized word usage;

an identification engine to;

apply a weighting mechanism to the different combinations of the output in the filtered set of content;

obtain a value corresponding to the different combinations of the output in the filtered set of content;

identify specialized language from the different combinations of the output in the filtered set of content, wherein the value corresponding to the different combinations of the output in the filtered set of content reaching at least a particular threshold indicates specialized language within that output; and

index the document based on the specialized language that is identified to identify other documents salient to the document based on the specialized language.

2. The system of claim 1, wherein the identification engine is to apply the weighting mechanism to the output produced by each of the multiple engines to obtain the value includes determining a relative density of text in a particular output compared to a density of the text in the document.

3. A method comprising:

receiving an output from multiple engines, wherein each engine of the multiple engines is to produce an output representative of a summary of a document specific to that engine based on a different type of engine selected from a group of engines comprising an extractive type of engine, an abstractive type of engine, and a frequency type of engine;

generating a filtered set of content in a single output to reduce a size of the output produced by the multiple engines, wherein the filtered set of content comprises different combinations of the output from the multiple engines that have different densities of specialized word usage;

applying a weighting mechanism to the different combinations of the output in the filtered set of content to obtain a value corresponding to each one of the different combinations of the output in the filtered set of content;

identifying jargon when the value corresponding to the each one of the different combinations of the output in the filtered set of content reaches at least a particular threshold; and

indexing the document based the jargon that is identified to identify other documents salient to the document based on the jargon.

4. The method of claim 3, wherein the multiple engines include multiple types of summarizer engines.

5. The method of claim 3, wherein the identifying the jargon is considered consistent when the output from the multiple engines is similar.

6. The method of claim 3, further comprising:

determining a domain of the document prior to application of the weighting mechanism, wherein the domain affects a type of weighting mechanism for application to the output specific to that engine.

7. A non-transitory machine-readable storage medium comprising instructions that when executed by a processing resource cause a computing device to:

receive an output from multiple engines, wherein each engine of the multiple engines is to produce an output representative of a summary of a document specific to that engine based on a different type of engine selected from a group of engines comprising an extractive type of engine, an abstractive type of engine, and a frequency type of engine;

generate a filtered set of content in a single output to reduce a size of the output produced by the multiple engines, wherein the filtered set of content comprises different combinations of the output from the multiple engines that have different densities of specialized word usage;

apply a weighting mechanism to the different combinations of the output in the filtered set of content to obtain a value corresponding to each one of the different combinations of the output in the filtered set of content;

compare the values of the output from the multiple engines;

identify jargon based on the comparison of values; and

index the document based the jargon that is identified to identify other documents salient to the document based on the jargon.

8. The non-transitory machine-readable storage medium of claim 7, wherein to identify the jargon based on the comparison of values comprises instructions that when executed by the processing resource causes the computing device to:

determine if the value corresponding to that output reaches at least a particular threshold.

9. The non-transitory machine-readable storage medium of claim 7, wherein to apply the weighting mechanism to the output specific to that engine to obtain the value corresponding to that output comprises instructions that when executed by the processing resource causes the computing device to:

determine a relative density of text in the output specific to the engine compared to a density of the text in the document.

10. The non-transitory machine-readable storage medium of claim 7, wherein the engine is to produce output representative of the summary of the document specific to the that engine wherein the output produced by that engine depends on a type of engine.