Context-based disambiguation of acronyms and abbreviations

• US 9,020,805 B2
• Filed: 09/23/2011
• Issued: 04/28/2015
• Est. Priority Date: 09/29/2010
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Boguraev, Branimir K., Chu-Carroll, Jennifer, Ferrucci, David A., Levas, Anthony T., Prager, John M.
• Primary Examiner(s): He, Jialong
• Application Number:
  US13/243,864
• Publication Number:
  US 20120084076A1
• Time in Patent Office: 1,313 Days
• Field of Search: 704/9
• US Class Current: 704/9
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06F 40/242 Dictionaries
• CPC Class Codes: G06F 40/274 Converting codes to words; Guess-ahead of partial word inputs
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

Context-based disambiguation of acronyms and/or abbreviations may determine a target abbreviation and one or more keywords appearing in context with the target abbreviation in a received passage, the target abbreviation representing a shortened form of one or more word. A contextual search query including the target abbreviation and said one or more keywords may be generated. A pseudo document index may be searched for one or more expansions of the target abbreviation by invoking the contextual search query, the pseudo document index containing index of one or more pseudo documents, associated one or more abbreviations and associated context keywords. One or more pseudo documents associated with the target abbreviation may be returned based on the searching of the pseudo document index.

Claims

1. A method for context-based disambiguation of abbreviations, comprising:

determining a target abbreviation and one or more keywords appearing in context with the target abbreviation in a received passage, the target abbreviation representing a shortened form of one or more words;

generating automatically by a processor, a contextual search query comprising the target abbreviation and said one or more keywords;

searching, by the processor, a pseudo document index for one or more expansions of the target abbreviation by invoking the contextual search query, the pseudo document index containing index of one or more pseudo documents by titles, associated one or more abbreviations, and associated context keywords, wherein the titles are the expansions of the abbreviations contained in the pseudo documents respectively;

returning one or more target pseudo documents associated with the target abbreviation based on the searching of the pseudo document index; and

providing one or more expansions associated with the target abbreviation based on the returned one or more target pseudo documents,

wherein the determining the target abbreviation and one or more keywords appearing in context with the target abbreviation in a received passage comprises generating one or more features that capture lexical and syntactic properties of the passage, and recognizing said target abbreviation and said one or more keywords appearing in context with the target abbreviation in the received passage based on the captured lexical and syntactic properties.

2. The method of claim 1, wherein said one or more pseudo documents are built to include a name containing an expansion of an abbreviation and content containing the abbreviation and associated context keywords.

3. The method of claim 1, further including generating a ranked list or scored list or combination of both, of the returned one or more target pseudo documents associated with the target abbreviation.

4. The method of claim 1, wherein the target abbreviation includes an acronym.

5. The method of claim 1, wherein the one or more pseudo documents are built based on a dictionary of abbreviations and associated expansions.

6. The method of claim 1, wherein a pseudo document of said one or more pseudo documents is generated for an expansion in an abbreviation expansion dictionary, by extracting data from sources that contain language occurring with the expansion.

7. A method for context-based disambiguation of abbreviations, comprising:

generating an abbreviation expansion dictionary by identifying a set of abbreviations with associated potential expansions;

generating, by a processor, a pseudo document for each expansion identified in the abbreviation expansion dictionary, the pseudo document comprising an abbreviation, associated expansion and one or more words that occur with said abbreviation, the generated pseudo document having a title corresponding to the associated expansion of the abbreviation that the pseudo document contains, the pseudo document generated at least by extracting data from sources that contain language commonly occurring with the expansion;

generating a pseudo document index indexing said abbreviation and said associated expansion; and

generating a machine learning classification model by generating one or more features that capture lexical and syntactic properties of a received passage, and building the machine learning classification model for recognizing one or more target abbreviations and one or more target keywords appearing in context with the target abbreviation in the received passage.