Document abstraction system and method thereof

• US 5,963,969 A
• Filed: 05/08/1997
• Issued: 10/05/1999
• Est. Priority Date: 05/08/1997
• Status: Expired due to Fees

• Current Assignee: William A. Tidwell
• Original Assignee: William A. Tidwell
• Inventors: Tidwell, William A.
• Primary Examiner(s): Feild, Joseph H.
• Assistant Examiner(s): Young, John L.
• Application Number:
  US08/848,406
• Time in Patent Office: 880 Days
• Field of Search: 707/531, 707/501, 707/1, 707/536, 707/532, 707/515, 707/514, 707/5, 707/500, 704/4
• US Class Current: 715/206
• CPC Class Codes: G06F 16/345 Summarisation for human users
• CPC Class Codes: G06F 40/30 Semantic analysis

Abstract

A document abstract system and methodology produces an abstract from a text by identifying sentences within the text that contain numerical information, such as dates and numbers. The sentences with numerical information along with the first and last sentences of the document are copied for producing the abstract. The computer generated abstract preferably includes a list of proper nouns and adjectives and most common words in the document.

Claims

1. A method of producing an abstract for a text, said text comprising one or more sentences, each sentence comprising one or more words, said method comprising the computer-implemented steps of:

determining the first sentence of said text;

determining the last sentence of said text;

determining, among the remaining sentences of said text, at most ten abstract sentences containing numeric information; and

producing said abstract based on said first sentence, said at most ten abstract sentences, and said last sentence.

2. The method of claim 1, wherein the step of determining one or more abstract sentences containing numeric information, includes the steps of:

searching for said numeric information within said text; and

if said numeric information is found, establishing a sentence containing said numeric information as an abstract sentence.

3. The method of claim 2, wherein the step of searching for said numeric information within said text includes the step of searching for numerals within said text.

4. The method of claim 2, wherein the step of searching for said numeric information within said tex includes the step of searching for spelled out numbers within said text.

5. The method of claim 2, wherein the step of searching for said numeric information within said text includes the step of searching for spelled out months within said text.

6. The method of claim 2, wherein the step of searching for said numeric information within said text includes the step of searching for spelled out days of the week within said text.

7. The method of claim 1, further comprising the steps of:

determining one or more proper names in said text, wherein a proper name is a word whose initial letter is capitalized when not at the beginning of a sentence of said one or more sentences; and

producing said abstract based on said one or more proper names.

8. The method of claim 7, further comprising the steps of:

counting a number of occurrences for each proper name of said one or more proper names; and

when producing said abstract based on said proper name, producing said abstract based on the number of occurrences for said proper name.

9. The method of claim 1, further comprising the steps of:

counting a number of occurrences for each word within said text;

generating a list of words from said text and the number of occurrences thereof;

pruning said list of words to produce a pruned list of words;

determining a plurality of common words from said pruned list of words, said plurality of common words comprising words from said pruned list of words having at least as many occurrences as any words from said pruned list of words not part of said plurality of common words; and

producing said abstract based on said plurality of common words.

10. The method of claim 9, wherein the step of pruning said list of words to produce a pruned list of words includes the step of removing a plurality of non-content words from said list of words, wherein said plurality of non-content words comprises pronouns, conjunctions, articles, and prepositions.

11. The method of claim 1, further comprising the steps of:

searching for a plurality of predetermined keywords within said text;

if one of said plurality of predetermined keywords is found, establishing a sentence containing said one of said plurality of predetermined keywords as a keyword abstract sentence; and

producing said abstract based on said first sentence, said one or more abstract sentences, said keyword abstract sentence, and said last sentence.

12. A method of producing an abstract for a text, said text comprising one or more sentences, each sentence comprising one or more words, said method comprising the computer-implemented steps of:

determining the first sentence of said text;

determining the last sentence of said text;

determining, among the remaining sentences of said text, one or more abstract sentences containing numeric information; and

counting a number of occurrences for each word within said text;

generating a list of words from said text and the number of occurrences thereof;

pruning said list of words to produce a pruned list of words including the step of removing proper names from said list of words, wherein a proper name has an initial letter that is capitalized when not at the beginning of a sentence of said one or more sentences;

determining a plurality of common words from said pruned list of words, said plurality of common words comprising words from said pruned list of words having at least as many occurrences as any words from said pruned list of words not part of said plurality of common words; and

producing said abstract based on said first sentence, said at most ten abstract sentences, said last sentence, and said plurality of common words.

13. A method of searching a plurality of texts, each comprising one or more sentences, said method comprising the computer-implemented steps of:

producing a plurality of abstracts, each abstract based on a corresponding text of said plurality of texts;

receiving a search key from a user;

searching said plurality of abstracts for an abstract that matches said search key; and

outputting said abstract that matches said search key;

wherein said abstract contains one or more sentences of the corresponding text, consisting essentially of the first sentence of the corresponding text, at most ten abstract sentences, each containing numeric information, and the last sentence of the corresponding text.

14. A method of searching a plurality of texts, each comprising one or more sentences, said method comprising the computer-implemented steps of:

receiving a search key from a user;

searching said plurality of texts for a text that matches said search key;

producing an abstract, based on a corresponding document of said plurality of texts that matches said search key; and

outputting said abstract;

wherein said abstract contains one or more sentences of the corresponding texts, consisting essentially of the first sentence of the corresponding text, at most ten abstract sentences, each containing numeric information, and the last sentence of the corresponding text.

15. A document abstraction system, configured to produce an abstract from a text, comprising one or more sentences each sentence comprising one or more words, said document abstraction system comprising:

means for determining the first sentence of said text;

means for determining the last sentence of said text;

means for determining, among the remaining sentences of said text, at most ten abstract sentences containing numeric information; and

means for producing said abstract based on said first sentence, said one or more abstract sentences, and said last sentence.

16. The system of claim 15, wherein said means for determining said one or more abstract sentences containing numeric information includes:

means for searching for said numeric information within said text; and

means for establishing a sentence containing said numeric information as an abstract sentence, if said numeric information is found.

17. The system of claim 15, further comprising:

means for determining one or more proper names in said text, wherein a proper name is a word whose initial letter is capitalized when not at the beginning of a sentence of said one or more sentences; and

means for producing said abstract based on said one or more proper names.

18. The system of claim 15, further comprising:

means for counting a number of occurrences for each word within said text;

means for generating a list of words from said text and the number of occurrences thereof;

means for pruning said list of words to produce a pruned list of words;

means for determining a plurality of common words from said pruned list of words, said plurality of common words comprising words from said pruned list of words having at least as many occurrences as any words from said pruned list of words not part of said plurality of common words; and

means for producing said abstract based on said plurality of common words.

19. The system of claim 15, further comprising:

means for searching for a plurality of predetermined keywords within said text;

means for establishing a sentence containing said one of said plurality of predetermined keywords as a keyword abstract sentence, if one of said plurality of predetermined keywords is found; and

means for producing said abstract based on said first sentence, said one or more abstract sentences, said keyword abstract sentence, and said last sentence.

20. A computer-readable medium having stored thereon sequences of instructions for producing an abstract from a document, comprising one or more sentences each comprising one or more words, said sequences of instructions includes sequences of instructions which, when executed by a processor, cause said processor to perform the steps of:

determining the first sentence of said text;

determining the last sentence of said text;

determining, among the remaining sentences of said text, at most ten abstract sentences containing numeric information; and

producing said abstract based on said first sentence, said one or more abstract sentences, and said last sentence.

21. The computer-readable medium of claim 20, wherein the step of determining one or more abstract sentences containing numeric information includes the steps of:

searching for said numeric information within said text; and

establishing a sentence containing said numeric information as an abstract sentence, if said numeric information is found.

22. The computer-readable medium of claim 20, further comprising sequences of instructions for performing the steps of:

determining one or more proper names in said text, wherein a proper name is a word whose initial letter is capitalized when not at the beginning of a sentence of said one or more sentences; and

producing said abstract based on said one or more proper names.

23. The computer-readable medium of claim 20, further comprising sequences of instructions for performing the steps of:

counting a number of occurrences for each word within said text;

generating a list of words from said document and number of occurrences thereof;

pruning said list of words to produce a pruned list of words;

determining a plurality of common words from said pruned list of words, said plurality of common words comprising words from said pruned list of words having at least as many occurrences as any words from said pruned list of words not part of said plurality of common words; and

producing said abstract based on said plurality of common words.

24. The computer readable medium of claim 20, further comprising sequences of instructions for performing the steps of:

searching for a plurality of predetermined keywords within said text;

if one of said plurality of predetermined keywords is found, establishing a sentence containing said one of said plurality of predetermined keywords as a keyword abstract sentence; and

producing said abstract based on said first sentence, said one or more abstract sentences, said keyword abstract sentence, and said last sentence.