Document retrieval method and document retrieval system

A document retrieval method for a document retrieval system that retrieves documents containing a specified string out of a set of documents registered beforehand, said document retrieval method comprising the steps of:

• extracting an n-gram, which is a sub-string having a predetermined length from a specified search term;
• extracting index data for said n-gram including text identification information, occurrence position information which indicates that the string is located at the n-th position from the beginning of said text data, top-of-word boundary information, and end-of-word boundary information;
• and
• searching documents without omission containing said specified search term using at least one of the group consisting of said text identification information, said occurrence position information, and top-of-word boundary information or end-of-word boundary information from an extracted index, while excluding, as “noise”, documents containing said specified search term as a subset of a different word.

A document retrieval system for retrieving documents containing a specified string out of a set of documents registered beforehand, said document retrieval system comprising:

• means for identifying word boundaries performing extraction of words from a text data of a document being registered, identifying top-of-word and end-of-word positions within said text data, and creating text data to which are added top-of-word and end-of-word identification information to top-of-word characters and end-of-word characters respectively;
• means for extracting n-grams for registration extracting sub-strings of a predetermined length (n-gram) from said created text data;
• means for registering and creating n-gram indexes with word boundary information extracting character position information for said extracted n-gram, including text identification information for said text data and occurrence position information which indicates that the string is located at the n-th position from the beginning of said text data, using identification results from said word boundary identification step to extract top-of-word boundary information if a leading character of said n-gram is at a top-of-word position and to extract end-of-word boundary information if a trailing character of said n-gram is at an end-of-word position, and creating and storing a search index corresponding to said extracted n-gram including information extracted for said n-gram.

A document retrieval system for retrieving documents containing a specified string out of a set of documents registered beforehand, said document retrieval system comprising:

• means for extracting n-grams for searches extracting sub-strings (n-grams) of a predetermined length from a specified search term;
• means for extracting indexes for searches extracting index data including text identification information, occurrence position information which indicates that the string is located at the n-th position from the beginning of said text data, top-of-word boundary identification information and end-of-word boundary identification information for said n-gram;
• means for analyzing search conditions analyzing a specified search condition to determine if said specified search condition is for a free-word search performing a search without omissions of documents containing said specified search term or is for a word boundary discrimination search that considers word boundaries and excludes a document if said specified search term exists in said document as a sub-string within a different word;
• means for performing free-word searches searching all documents containing said specified search term without omission based on said text identification information and occurrence position information in an index extracted by said search index extracting means if the result of said analysis is a free-word search;
• and
• means for performing word-discrimination searches searching all documents containing said specified search term without omission while excluding, as “noise”, documents containing said specified search string as a substring of a different word if the result of said analysis is a word-discrimination search, said searching being performed using at least one of the group consisting of text identification information, occurrence position information, and top-of-word boundary identification information or end-of-word boundary identification information from an index extracted by said search index extracting means.