Method for language-independent text tokenization using a character categorization

• US 4,991,094 A
• Filed: 04/26/1989
• Issued: 02/05/1991
• Est. Priority Date: 04/26/1989
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Gunther, Michael D., Fagan, Joel L., Zamora, Elena M., Tsao, Chien C., Passon, Greg, Over, Paul D., Zamora, Antonio
• Primary Examiner(s): Fleming, Michael R.
• Assistant Examiner(s): Sheikh, Ayaz R.
• Application Number:
  US07/344,341
• Time in Patent Office: 650 Days
• Field of Search: 364/419, 364/900, 364/200, 364/
• US Class Current: 704/9
• CPC Class Codes: G06F 40/284 Lexical analysis, e.g. tokenisation or collocates

Abstract

A computer method is disclosed to isolate linguistically salient strings ("words") from a natural language text stream. The process is applicable to a variety of computer hardware, to any character encoding scheme, and to the idiosyncrasies of most natural languages.

Claims

1. A computer method for isolating words from an input stream of characters forming natural language text, the method comprising the steps of:

storing in a data processor a character classification table which describes each character in a character code as being either a delimiter character, an alpha/numeric character or a conditional delimiter character which assumes a function of a delimiter character when it occurs in predefined character contexts;

inputting to said data processor an input stream of characters which are members of said character code, the input stream forming natural language text;

building in said data processor a string of alpha/numeric characters to form a word from said input stream of characters produced by said inputting step;

isolating in said data processor three consecutive characters from said inputting step of said input stream, as a previous character, a current character and a next character;

accessing in said data processor said character classification table in response to said isolating step, to determine if said current character is a delimiter character, an alpha/numeric character or a conditional delimiter character;

appending in said data processor said current character to said string when said character classification table identifies it as an alpha/numeric character in said accessing step;

signaling an output signal from said data processor that said string is a complete word when said character classification table identifies said current character as a delimiter character in said accessing step;

analyzing in said data processor said previous character, said current character and said next character to determine if said current character assumes the function of a delimiter character when said character classification table identifies said current character as a conditional delimiter character in said accessing step.

2. The computer method of claim 1, wherein said analyzing step further comprises:

appending in said data processor said current character to said string when said previous character is an alphabetic character and said next character is an alphabetic character.

3. The computer method of claim 1, wherein said analyzing step further comprises:

appending in said data processor said current character to said string when said previous character is a numeric character and said next character is a numeric character.

4. The computer method of claim 1, wherein said analyzing step further comprises:

appending in said data processor said current character to said string when said previous character is not an alphabetic character and not a numeric character and said next character is a numeric character.

5. The computer method of claim 1, wherein said analyzing step further comprises:

appending in said data processor said current character to said string when said previous character is not an alphabetic character and not a numeric character and said current character is a same as said previous character.

6. The computer method of claim 1, which further comprises:

storing in said data processor a second character classification table which describes each character in a second character code as being either a delimiter character, an alpha/numeric character or a conditional delimiter character which assumes the function of a delimiter character when it occurs in predefined character contexts;

selecting in said data processor either the first said character code or said second character code for said input character stream;

selectively accessing in said data processor in said accessing step, either the first said character classification table or said second character classification table, in response to said step of selecting either said first character code or said second character code.

7. The computer method of claim 6, wherein said first said classification table characterizes a first natural language and said second classification table characterizes a second natural language.

8. The computer method of claim 6, wherein said first said classification table characterizes a first type data processing code and said second classification table characterizes a second type data processing code.

9. The computer method of claim 1, which further comprises:

storing in said data processor a second character classification table which describes each character in a second character code as being either a delimiter character, an alpha/numeric character or a conditional delimiter character which assumes the function of a delimiter character when it occurs in predefined character contexts, the first said classification table characterizing a first natural language and said second classification table characterizing a second natural language;

selectively accessing in said accessing step, either the first said character classification table or said second character classification table, in response to selecting either said first natural language or said second natural language for said input character stream.

10. The computer method of claim 1, which further comprises the step of comparing in said data processor said characters in said input stream with a predefined set of prefixes and suffixes to indicate additional locations of words in said input stream.

11. The computer method of claim 1, which further comprises the steps of:

outputting from said data processor said complete word identified in said signaling step, as a search term and inputting it into an index for dictionary accessing.

12. The computer method of claim 1, which further comprises the steps of:

outputting from said data processor said complete word identified in said signaling step, as a search term and inputting it into an index for database accessing.

13. The computer method of claim 1, which further comprises the steps of:

outputting from said data processor said complete word identified in said signaling step, as a word in a first natural language and inputting it into a language translator for translation into a word in a second natural language.

14. The computer method of claim 1, applied to an input stream containing chemical literature terms.

15. The computer method of claim 1, applied to an input stream containing mathematical literature terms.