Domain specific representation of document text for accelerated natural language processing

• US 9,898,447 B2
• Filed: 06/22/2015
• Issued: 02/20/2018
• Est. Priority Date: 06/22/2015
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Desai, Rajesh M., Housfater, Alon S., Parker, Philip E., Raphael, Roger C.
• Primary Examiner(s): ZHU, RICHARD Z
• Application Number:
  US14/746,327
• Publication Number:
  US 20160371255A1
• Time in Patent Office: 974 Days
• CPC Class Codes: G06F 16/22 Indexing; Data structures therefor; Storage structures
• CPC Class Codes: G06F 16/2228 Indexing structures
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: G06F 16/367 Ontology
• CPC Class Codes: G06F 16/901 Indexing; Data structures therefor; Storage structures for retrieval from the web G06F16/951
• CPC Class Codes: G06F 40/151 Transformation
• CPC Class Codes: G06F 40/242 Dictionaries
• CPC Class Codes: G06F 40/284 Lexical analysis, e.g. tokenisation or collocates
• CPC Class Codes: G06F 9/00 Arrangements for program control, e.g. control units program control for peripheral devices G06F13/10
• CPC Class Codes: G06F 9/46 Multiprogramming arrangements
• CPC Class Codes: G06T 1/20 Processor architectures; Processor configuration, e.g. pipelining

Abstract

Provided are techniques for a domain specific representation of document text for accelerated natural language processing. A document is selected from a set of documents to be analyzed. A character stream from the document is converted into a token stream based on tokenization rules. Irrelevant tokens are removed from the token stream. The tokens remaining in the token stream are converted into an integer domain representation based on a domain specific ontology dictionary. The integer domain representation are stored to a Graphics Processing Unit (GPU) processing queue of each of one or more GPUs. Then, a result set is received from the one or more GPUs.

Claims

1. A computer program product, the computer program product comprising a computer readable storage medium having program code embodied therewith, the program code executable by one of a Central Processing Unit (CPU) processor and at least one Graphics Processing Unit (GPU) processor to perform:

selecting a document from a set of documents to be analyzed;

converting a character stream from the document into a token stream based on tokenization rules;

removing irrelevant tokens from the token stream;

converting tokens remaining in the token stream into an integer domain representation based on a domain specific ontology dictionary, wherein each of the tokens is mapped to an integer based on mappings in the domain specific ontology dictionary;

storing the integer domain representation to a Graphics Processing Unit (GPU) processing queue of each of one or more GPUs;

receiving a result set from the one or more GPUs, wherein the result set includes tuples of

     1) a specific pattern and

     2) an offset from a beginning of the integer domain representation, and wherein the result set is generated using a compiled super Regular Expression (REGEX) that is compiled using the domain specific ontology dictionary;

storing the result set into an index for use in processing a search query; and

persisting the integer domain representation once for the document for processing of the document with a different compiled super REGEX.

2. The computer program product of claim 1, wherein the program code is executable by at least one processor to perform:

persisting to the domain specific ontology dictionary any newly discovered relevant integer domain representations for new tokens for the document that are relevant to a domain of the domain specific ontology dictionary.

3. The computer program product of claim 1, wherein a Software as a Service (SaaS) is configured to perform computer program product operations.

4. The computer program product of claim 1, wherein the program code is executable by at least one processor to perform:

selecting a REGEX rule from a set of character based domain REGEX rules;

applying the domain specific ontology dictionary to map a character based REGEX to an integer based REGEX;

compiling the integer based REGEX; and

adding the compiled integer based REGEX to the super REGEX.

5. The computer program product of claim 1, wherein the domain specific ontology dictionary is selected from a set of domain specific ontology dictionaries for different domains.

6. The computer program product of claim 1, wherein the domain specific ontology dictionary represents domain terminology as a fixed set of integers.

7. The computer program product of claim 1, wherein the program code is executable by at least one processor to perform:

assigning a group of tokens to a group, which is assigned one integer domain representation.

8. The computer program product of claim 1, wherein the program code is executable by at least one processor to perform:

parallelizing document level extraction using REGEX processing with GPU acceleration over integers in the integer domain representation.

9. A computer system, comprising:

one or more Central Processing Unit (CPU) processors and Graphics Processing Unit (GPU) processors, one or more computer-readable memories and one or more computer-readable, tangible storage devices; and

program instructions, stored on at least one of the one or more computer-readable, tangible storage devices for execution by at least one of the one or more CPU processors via at least one of the one or more memories, to perform operations comprising;

selecting a document from a set of documents to be analyzed;

converting a character stream from the document into a token stream based on tokenization rules;

removing irrelevant tokens from the token stream;

converting tokens remaining in the token stream into an integer domain representation based on a domain specific ontology dictionary, wherein each of the tokens is mapped to an integer based on mappings in the domain specific ontology dictionary;

storing the integer domain representation to a Graphics Processing Unit (GPU) processing queue of each of one or more GPUs;

receiving a result set from the one or more GPUs, wherein the result set includes tuples of

     1) a specific pattern and

     2) an offset from a beginning of the integer domain representation, and wherein the result set is generated using a compiled super Regular Expression (REGEX) that is compiled using the domain specific ontology dictionary;

storing the result set into an index for use in processing a search query; and

persisting the integer domain representation once for the document for processing of the document with a different compiled super REGEX.

10. The computer system of claim 9, wherein the operations further comprise:

persisting to the domain specific ontology dictionary any newly discovered relevant integer domain representations for new tokens for the document that are relevant to a domain of the domain specific ontology dictionary.

11. The computer system of claim 9, wherein a Software as a Service (SaaS) is configured to perform computer system operations.

12. The computer system of claim 9, wherein the operations further comprise:

selecting a REGEX rule from a set of character based domain REGEX rules;

applying the domain specific ontology dictionary to map a character based REGEX to an integer based REGEX;

compiling the integer based REGEX; and

adding the compiled integer based REGEX to the super REGEX.

13. The computer system of claim 9, wherein the domain specific ontology dictionary is selected from a set of domain specific ontology dictionaries for different domains.

14. The computer system of claim 9, wherein the domain specific ontology dictionary represents domain terminology as a fixed set of integers.

15. The computer system of claim 9, wherein the operations further comprise:

assigning a group of tokens to a group, which is assigned one integer domain representation.

16. The computer system of claim 9, wherein the operations further comprise:

parallelizing document level extraction using REGEX processing with GPU acceleration over integers in the integer domain representation.