Method for automatically selecting collections to search in full text searches

• US 5,845,278 A
• Filed: 09/12/1997
• Issued: 12/01/1998
• Est. Priority Date: 09/12/1997
• Status: Expired due to Term

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Inioseek Corporation
• Inventors: Chang, William I., Kirsch, Steven T.
• Primary Examiner(s): Black, Thomas G.
• Assistant Examiner(s): Mizrahi, Diane D.
• Application Number:
  US08/928,542
• Time in Patent Office: 445 Days
• Field of Search: 707/1, 707/5, 707/4, 707/3, 707/102, 707/103
• US Class Current: 707/688
• CPC Class Codes: G06F 16/30 of unstructured textual data document management systems G06F16/93
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06F 16/9538 Presentation of query results
• CPC Class Codes: Y10S 707/917 Text
• CPC Class Codes: Y10S 707/99931 Database or file accessing
• CPC Class Codes: Y10S 707/99933 Query processing, i.e. searching
• CPC Class Codes: Y10S 707/99934 Query formulation, input preparation, or translation
• CPC Class Codes: Y10S 707/99935 Query augmenting and refining, e.g. inexact access
• CPC Class Codes: Y10S 707/99943 Generating database or data structure, e.g. via user interface

Abstract

A method of selecting a subset of a plurality of document collections for searching in response to a predetermined query is based on accessing a meta-information data file that describes the query significant search terms that are present in a particular document collection correlated to normalized document usage frequencies of such terms within the documents of each document collection. By access to the meta-information data file, a relevance score for each of the document collections is determined. The method then returns an identification of the subset of the plurality of document collections having the highest relevance scores for use in evaluating the predetermined query. The meta-information data file may be constructed to include document normalized term frequencies and other contextual information that can be evaluated in the application of a query against a particular document collection. This other contextual information may include term proximity, capitalization, and phraseology as well as document specific information such as, but not limited to collection name, document type, document title, authors, date of publication, publisher, keywords, summary description of contents, price, language, country of publication, publication name. Statistical data for the collection may include such as, but not limited to number of documents in the collection, the total size of the collection, the average document size and average number of words in the base document collection.

Claims

1. A method of selecting a subset of a set of document collections containing documents to search based upon a predetermined query text including a search term, said method comprising the steps of:

a) accessing a meta-file representative of said set of document collections, including a search term occurrence list;

b) determining a document frequency term for said search term relative to each of said document collections within said set of document collections and an inverse collection frequency term for said set of document collections, said inverse collection frequency term being proportional to a ratio of the number of documents in said set of document collections and the number of documents in set of document collections that include said search term;

c) determining a term ranking for each of said document collections that is proportional to the respective said document frequency terms and said inverse collection frequency term;

d) selecting said subset of said set of document collections based on the relative term ranking of each of said document collections.

2. The method of claim 1 wherein said meta-file includes a list of search terms that occur in said set of document collections correlated to the collection and document of each search term'"'"'s occurrence.

3. The method of claim 2 wherein said predetermined query text includes a plurality of search terms, wherein said steps (a) through (c) are performed for each of said plurality of search terms, and where a final ranking for each of said document collections is determined as the sum of said term rankings for each of said plurality of search terms, said step (d)being performed based on the relative final ranking of each of said document collections.

4. The method of claim 3 wherein said document frequency terms and said inverse collection frequency term determined for each of said document collections is scaled to increase the numerical distribution of said document frequency terms while maintaining proportionality with said inverse collection frequency term.

5. The method of claim 4 wherein a log function is used to scale said document frequency terms and said inverse collection frequency term determined for each of said document collections.

6. The method of claim 2 or 4 wherein said meta-file includes a plurality of records, a predetermined record being representative of a respective document collection, said predetermined record including said search term occurrence list and a plurality of fielded terms provided to store characteristics of said respective document collection.

7. The method of claim 6 wherein said plurality of fielded terms may be categorized as storing statistical data and pre-search collection qualifying data.

8. The method of claim 7 wherein said statistical data may include the number of documents represented by said predetermined record and the number of occurrences of a predetermined search term that appears in said search term occurrence list, and wherein said pre-search collection qualifying data may include a language type identifier, a publication type identifier, a publication name, a publication date, and a cost identifier.

9. The method of claim 8 wherein said predetermined query text includes a post-search qualifying term, and wherein said step of selecting said subset of document collections includes a step of qualifying said subset of document collections selected based on said post-search qualifying term.

10. A method of selecting a subset of a set of document collections to search based upon a predetermined query text optionally including any of a search term, a pre-search qualifier, and a post-search qualifier, said method comprising the steps of:

a) accessing a meta-information data file that includes a plurality of records representing said set of document collections, each said document collection representing a plurality of documents;

b) pre-qualifying a set of said plurality of records based upon said pre-search qualifier, if any;

c) determining a search term frequency value for each of said pre-qualified set of said plurality of records with respect to said search term if any, said search term frequency values being normalized against a common factor representative of the frequency of qualifying occurrences of said search term within said documents of said pre-qualified set;

d) determining a search term ranking for each of said pre-qualified set of said plurality of records based upon said frequency values and said common factor; and

e) selecting said subset of said set of document collections to search based on said search term rankings and said post-search qualifier, if any.

11. The method of claim 10 wherein said document collections have a searchable collection index, and wherein said method further comprises the step of applying said search term against said searchable collection index to select one or more predetermined documents.

12. The method of claim 11 further including a step of optionally modifying any of said search term, said pre-search qualifier and said post-search qualifier and re-performing said steps (b) through (e).

13. The method of claim 12 wherein said normalized document usage frequencies for said predetermined document collection are determined in relation to the frequency of occurrence of said search significant terms within said predetermined document collection.

14. A method for selecting a subset of a set of document collections to search dependant on a predetermined query term, each said document collection including a plurality of documents, said set of document collections being represented as a meta-index that stores search terms and statistical data representative of said set of document collections and said document collections being represented by respective collection indexes that store search terms and statistical data representative of the documents within respective document collections, said method comprising the steps of:

a) determining a collection ranking for each said document collection with respect to said predetermined query term with reference to said meta-index, each said collection ranking being normalized with respect to the qualified occurrence of said predetermined query term within the documents of said set of document collections;

c) identifying said document collections within said subset of document collections potentially most relevant for searching based on said predetermined query term.

15. The method of claim 14 wherein said qualified occurrence of said predetermined query term is representable as a ratio of the total number of documents within said set of documents and the number of documents, in said set of document collections, that have at least one occurrence of said predetermined query term.

16. The method of claim 15 wherein each said collection ranking is dependant on the number of documents within a respective collection that have at least one occurrence of said predetermined query term relative to a qualified number of documents within said respective collection.

17. The method of claim 16 wherein said qualified number of documents within said respective collection is dependant on the number of documents within said respective collection subject to a minimum threshold number of documents.

18. The method of claim 17 wherein said collection ranking for each said collection is further dependant on the sum of the said collection rankings separately determined with respect to each of a set of query terms including said predetermined query term.

19. The method of claim 18 wherein said qualified occurrence of said predetermined query term is once per document and said minimum threshold number of documents is 100.

20. The method of claim 17 or 19 wherein said step of identifying said document collections includes selecting as said subset a predetermined number of said documents having the highest relative collection rankings from said set of document collections.