Providing capitalization correction for unstructured excerpts

• US 7,451,398 B1
• Filed: 11/18/2003
• Issued: 11/11/2008
• Est. Priority Date: 11/18/2003
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Rohrs, Christopher
• Primary Examiner(s): Bashore; William L
• Application Number:
  US10/716,951
• Time in Patent Office: 1,820 Days
• Field of Search: 715531-534, 715/530, 715/540, 715255-261, 715/271
• US Class Current: 715/271
• CPC Class Codes: G06F 40/232 Orthographic correction, e.g. spell checking or vowelisation

Abstract

Providing capitalization correction for unstructured excerpts is described. An excerpt of unstructured content is tokenized into a set of words. The set of words is analyzed for correct capitalization. Individual characters constituting at least one such word in the set of words are evaluated. The at least one such word is skipped if determined to be of a predefined type.

Claims

1. A computer system for building a lexicon for use in capitalization correction for unstructured excerpts, comprising:

at least one processing unit;

at least one storage device being coupled with the at least one processing unit and storing program code;

a ripper adapted to assemble a list of word sets from unstructured content, at least one of the word sets comprising a word and at least two non-standard capitalization variations for the word; and

an aggregator adapted to aggregate at least one of the word sets, the aggregator including

an analyzer adapted to identify non-standard capitalization variations based on at least one criteria; and

a non-standard capitalization selector adapted to select at least one of the identified non-standard capitalization variations within one of the at least one word sets, and adding the selected at least one of the identified non-standard capitalization variations to the lexicon, wherein the lexicon includes records, each record including a word, wherein the lexicon is indexed by the words included in the records, and wherein at least one of the records includes more than one non-standard capitalization variation.

2. A computer system according to claim 1, further comprising:

a tokenizer adapted to tokenize the excerpt into the one or more words and one or more punctuation marks.

3. A computer system according to claim 2, wherein hyphenated words are split into a plurality of the words.

4. A computer system according to claim 1, wherein at least one of the non-standard capitalization variations occurs in an excerpt having fewer than half of individual letters provided in uppercase.

5. A computer system according to claim 1, further comprising:

a normalizer adapted to normalize a plurality of the words extracted relative to a source of the unstructured excerpt.

6. A computer system according to claim 1, wherein non-standard capitalization variations that are identified based on one or more criteria comprise only those non-standard capitalization variations having at least four occurrences.

7. A computer system according to claim 1, wherein at least one of the non-standard capitalization variations has any individual letter other than the first individual letter provided in uppercase.

8. A computer system according to claim 1, further comprising:

a validator adapted to apply implicit rules for capitalization, and skipping each of the non-standard capitalization variations subject to at least one such implicit rule.

9. A computer system according to claim 8, wherein the implicit rules comprise skipping each of the non-standard capitalization variations based on position within a sentence or phrase.

10. A computer system according to claim 8, wherein the implicit rules comprise at least one of (A) the non-standard capitalization variation being a number, (B) the non-standard capitalization variation having no vowels, and (C) the non-standard capitalization variation constituting at least one of an article, conjunction and preposition.

11. A computer system according to claim 8, wherein the implicit rules comprise normalizing a number of occurrences for each of the non-standard capitalization variations relative to a source of the non-standard capitalization variations.

12. A computer system according to claim 8, wherein each of the word sets includes a word and at least one non-standard capitalization variation, each of the at least one non-standard capitalization variation including a frequency of occurrence count.

13. A computer system according to claim 1, further comprising:

a hash table maintaining the lexicon.

14. A computer system according to claim 13,

wherein the hash table is indexed by words.

16. The computer-implemented method of claim 14 wherein a non-standard capitalization variation is identified using the at least one criteria only if it occurs at least four times in the text corpus.

17. The computer-implemented method of claim 14 further comprising:

d) accepting a word having a capitalization defining which, if any, of the characters of the word are capitalized; and

e) performing a capitalization correction function on the word using the generated lexicon.

18. The computer-implemented method of claim 17 wherein the act of performing a capitalization correction function includes

determining if the capitalization of the word matches a capitalization variation in the lexicon, and

not changing the capitalization of the word if it was determined to match a capitalization variation in the lexicon.

19. The computer-implemented method of claim 17 wherein the act of performing a capitalization correction function includes

determining if the capitalization of the word matches a non-standard capitalization variation in the lexicon, which non-standard capitalization variation meets a frequency criteria, and

not changing the capitalization of the word if it was determined to match a non-standard capitalization variation in the lexicon.

15. A computer-implemented method comprising:

a) generating a plurality of word sets from a text corpus, at least one of the words sets including

a word identified from the text corpus,

at least one non-standard capitalization variation of the word included in the word set, and

a frequency of occurrence of each of the at least one non-standard capitalization variation of the word included in the word set;

b) generating a lexicon using the generated plurality of word sets, wherein the lexicon includes, for each of a plurality of words, at least one capitalization variation identified using at least one criteria, wherein at least one of the words of the lexicon includes more than one non-standard capitalization variation identified using the at least one criteria; and

c) storing the generated lexicon.

20. Apparatus comprising:

a) means for generating a plurality of word sets from a text corpus, at least one of the word sets including

a word identified from the text corpus,

at least one non-standard capitalization variation of the word included in the word set, and

a frequency of occurrence of each of the at least one non-standard capitalization variation of the word included in the word set; and

b) means for generating a lexicon using the generated plurality of word sets, wherein the lexicon includes, for each of a plurality of words, at least one capitalization variation identified using at least one criteria, wherein at least one of the words of the lexicon includes more than one non-standard capitalization variation identified using the at least one criteria.

21. The apparatus of claim 20 wherein a non-standard capitalization variation is identified using the at least one criteria only if it occurs at least four times in the text corpus.

22. The apparatus of claim 20 further comprising:

c) means for accepting a word having a capitalization defining which, if any, of the characters of the word are capitalized; and

d) means for performing a capitalization correction function on the word using the generated lexicon.

23. The apparatus of claim 22 wherein the means for performing a capitalization correction function

determine if the capitalization of the word matches a capitalization variation in the lexicon, and

do not change the capitalization of the word if it was determined to match a capitalization variation in the lexicon.

24. The apparatus of claim 22 wherein the means for performing a capitalization correction function

determine if the capitalization of the word matches a capitalization variation in the lexicon, which capitalization variation meets a frequency criteria, and

do not change the capitalization of the word if it was determined to match a capitalization variation in the lexicon.