Method for categorizing documents into subjects using relevance normalization for documents retrieved from an information retrieval system in response to a query

A method for storing input information in an information retrieval system database, said input information being formed of a collection of English language words, comprising the steps of:

• (A) determining an information subject category within said information retrieval system database;
• (B) determining a subject lexicon corresponding to said information subject category and containing information representative of said information subject category, said subject lexicon containing a plurality of classifier words, wherein each of said plurality of classifier words contains a first English language word (a) and a second English word (b), and each of said plurality of classifier words has a discriminator weight (w(a, b)) calculated in accordance with the formula:
• w(a,b)=log (P(ab)/P(a)P(b))
• wherein P(a) is a probability that said first English language word (a) occurs in a document pool, P(b) is a probability that said second English language word (b) occurs in said document pool, and P(ab) is a probability that said first English language word (a) is positioned adjacent to said second English language word (b) in a subject corpus;
• (C) comparing said collection of English language words from said input information with said classifier words from said subject lexicon;
• and
• (D) storing said input information in said information subject category in accordance with said comparing of step (C).

A subjector for selectively storing input information in an information retrieval system database, said input information being formed of a collection of English language words, comprising:

• at least one information subject category within said information retrieval system database;
• a subject lexicon corresponding to said information subject category and containing information representative of said information subject category, said subject lexicon containing a plurality of classifier words, wherein each of said plurality of classifier words contains a first English language word (a) and a second English word (b), and each of said plurality of classifier words has a discriminator weight (w(a, b)) calculated in accordance with the formula:
• w(a, b)=log (P(ab)/P(a)P(b))
• wherein P(a) is a probability that said first English language word (a) occurs in a document pool, P(b) is a probability that said second English language word (b) occurs in said document pool, and P(ab) is a probability that said first English language word (a) is positioned adjacent to said second English language word (b) in a subject corpus;
• an information comparing unit for first comparing said collection of English language words from said input information with said classifier words from said subject lexicon;
• and
• memory for storing said input information in said information subject category in accordance with said first comparing.

A method for storing input information in an information retrieval system database, said input information being formed of a collection of English language words, comprising the steps of:

• (a) determining an information subject category within said information retrieval system database;
• (b) determining a first subject lexicon corresponding to said information subject category and containing information representative of said information subject category, said first subject lexicon containing a plurality of classifier words, wherein generally all of said classifier words in said first subject lexicon contain at least one English language word, said step of determining said first subject lexicon including:
• (i) first selecting a first plurality of subjected documents from said information retrieval system database relating to said information subject category to provide first selected documents;
• (ii) determining a preliminary lexicon in accordance with said first plurality of subjected documents;
• (iii) comparing selected documents with said preliminary lexicon;
• (iv) second selecting a second plurality of documents in accordance with said comparing of step (iii);
• (v) determining whether documents of said second plurality of documents belong in said first subject category:
• (vi) removing documents from said second plurality of documents in accordance with the determining of step (v) to provide a third plurality of documents;
• and
• (vii) determining said first subject lexicon in accordance with said third plurality of documents;
• (c) comparing said collection of English language words from said input information with said classifier words from said first subject lexicon, wherein said collection of English language words from said input information form at least one document;
• and
• (d) storing said input information in said information subject category in accordance with said comparing of step (c).