Michael L Barbarino

US Patent:

55440495, Aug 6, 1996

Filed:

May 22, 1995

Appl. No.:

8/447317

Inventors:

Richard D. Henderson - San Jose CA
Michael J. Barbarino - Moss Beach CA

Assignee:

Xerox Corporation - Stamford CT

International Classification:

G06F 1730

US Classification:

36441919

Abstract:

A method for performing a search of a plurality of documents for similarity to a query word includes retrieving a first document, and determining a number of occurrences of the at least one query word in the first document. Then, a next document is retrieved and a number of occurrences of the at least one query word in the next document is determined. The steps are repeated until each of the plurality of documents have been retrieved, and the number of occurrences of the at least one query word has been determined in each of the plurality of documents. The query word can include a plurality of query words, all of which are searched in each document, in turn, rather than being searched word by word in the whole collection of documents. The documents are then ranked according to the number of occurrences of the query words determined in each document, and a list of documents is produced according to the document ranking.

US Patent:

55420907, Jul 30, 1996

Filed:

Jul 27, 1994

Appl. No.:

8/280963

Inventors:

Richard D. Henderson - San Jose CA
Michael J. Barbarino - Moss Beach CA

Assignee:

Xerox Corporation - Stamford CT

International Classification:

G06F 1730

US Classification:

395600

Abstract:

A method for searching a document corpus for query terms includes generating a list of document terms including a term signature for each term based upon characteristics of a number of adjacent terms. The term signatures can be generated by generating a bit vector for each term within a predetermined adjacent number of terms from each document term, such as through application of a hash function. The bit vectors can then be combined to form the term signature. The word signature alternatively can be generated using one or more morphological properties of the terms. The predetermined adjacent number of terms can be the number of search terms minus one, and may precede, follow, or both precede and follow the document term for which the term signature is generated. A search signature is generated for the query terms excluding a reference term, based upon the predetermined characteristics. The term signature of the reference term is compared with the search signature, and an indication is provided when the term signature of the reference term and the search signature match.