Document retrieval device

Claim 2

A document retrieval device that classifies a group of documents, each document having structural elements arranged in a logical hierarchical format, each structural element including at least one of a heading and a content, the documents stored in a document storage device, the retrieval device comprising: logical structure analysis means for analyzing the logical hierarchical format of the documents and for obtaining structural elements and a hierarchical relationship between the structural elements within each document; classification unit designation means for designating a classification unit showing which level the structural elements to be classified are at within the hierarchical relationship; fundamental vector generation means for extracting a key word from the content of each structural element of the classification unit that is designated by the classification unit designation means, and for generating a fundamental vector based on the extracted key word; heading vector generation means for extracting a key word from the heading of each structural element that is superordinate to the structural element used to generate the fundamental vector within said hierarchical relationship, and for generating a heading vector based on the extracted key word; vector synthesis means for generating, for each structural element of the classification unit, a composite vector based on the corresponding fundamental and heading vectors; and classification means for calculating a degree of similarity among the composite vectors and for classifying the structural elements of the documents based on the degree of similarity.

Claim 3

The document retrieval device of claim 1, further comprising; classification result maintenance means for maintaining the classification results; and reclassification designation means for reclassifying the structural elements based on the results maintained in the classification result maintenance means.

Claim 4

The document retrieval device of claim 1, further comprising: a display element selection section that selects the structural element of the classification unit from the result of the classifications generated by the classification means; a display range selection section that allows selection of a range of structural element to be displayed from the structural elements that include the classification unit; and a display element extraction section that extracts from the document storage device the structural element selected by the display element selection section, and that displays the range of the structural element selected with said display range selection section.

Claim 5

A document retrieval device that retrieves structural elements of documents, each document having structural elements arranged in a logical hierarchical format, each structural element including at least one of a heading and a content, the documents stored in a document storage device, the retrieval device comprising: logical structure analysis means for analyzing the logical hierarchical format of said documents and for obtaining structural elements and a hierarchical relationship between the structural elements within each document; retrieval unit designation means for designating a retrieval unit, the retrieval unit showing which level the structural elements to be retrieved are at within the hierarchical relationship; fundamental vector generation means for extracting a key word from the content of each structural element of the retrieval unit designated by the retrieval unit designation means, and for generating a fundamental vector based on the extracted key word; heading vector generation means for extracting a key word from the heading of each structural element that is superordinate to the structural element used to generate the fundamental vector within said hierarchical relationship, and for generating a heading vector based on the extracted key word; vector synthesis means for generating, for each structural element of the retrieval unit, a composite vector based on the corresponding fundamental and heading vectors; query input means for inputting a query; query vector generation means for generating a query vector from the query; and retrieval means for calculating a degree of similarity between the query vector and each composite vector and for retrieving the structural elements of the retrieval unit having a predetermined degree of similarity relative to composite vectors.

Claim 6

A document retrieval device that retrieves structural elements of documents, each document having structural elements arranged in a logical hierarchical format, each structural element including at least one of a heading and a content, the documents stored in a document storage device, the retrieval device comprising: logical structure analysis means for analyzing the logical hierarchical format of said documents and for obtaining structural elements and a hierarchical relationship between the structural elements within each document; retrieval unit designation means for designating a retrieval unit, the retrieval unit showing which level the structural elements to be retrieved are at within the hierarchical relationship; fundamental vector generation means for extracting a key word from the content of each structural element of the retrieval unit designated by the retrieval unit designation means, and for generating a fundamental vector based on the extracted key word; heading vector generation means for extracting a key word from the heading of each structural element that is superordinate to the structural element used to generate the fundamental vector within said hierarchical relationship, and for generating a heading vector based on the extracted key word; vector synthesis means for generating, for each structural element of the retrieval unit, a composite vector based on the corresponding fundamental and heading vectors; query input means for inputting a query; query vector generation means for generating a query vector from the query; and retrieval means for calculating a degree of similarity between the query vector and each composite vector and for retrieving the structural elements of the retrieval unit having a predetermined degree of similarity relative to composite vectors.

Claim 7

The document retrieval device of claim 4, wherein the query is a key word list or a natural language text.

Claim 8

The document retrieval device of claim 4, further comprising: a display element selection section that selects a structural element to be displayed from the structural elements retrieved by the retrieval means; a display range selection section that allows selection of a range of the structural element to be displayed from the structural elements that include the retrieval unit; and a display element extraction section that extracts from the document storage device the structural element selected by said display range selection section and that displays the structural element on a display means.

Claim 9

A document retrieval device that classifies a group of documents, each document having structural elements arranged in a logical hierarchical format, each structural element having a content, the documents stored in a document storage device, the retrieval device comprising: logical structure analysis means for analyzing the logical hierarchical format of the documents and for obtaining structural elements and a hierarchical relationship between the structural elements within each document; classification unit designation means for designating a classification unit showing which level the structural elements to be classified are at within the hierarchical relationship; fundamental vector generation means for extracting a key word from the content of each structural element of the classification unit that is designated by the classification unit designation means, and for generating a fundamental vector based on the extracted key word; content vector generation means for extracting a key word from the content of each structural element other than the structural element used to generate the fundamental vector, and for generating a content vector for each structural element based on the extracted key word; vector synthesis means for generating, for each structural element of the classification unit, a composite vector based on the corresponding fundamental and content vectors; and classification means for calculating a degree of similarity among the composite vectors and for classifying the structural elements of the documents based on the degree of similarity.

Claim 10

A document retrieval device that classifies a group of documents, each document having structural elements arranged in a logical hierarchical format, each structural element having a content, the documents stored in a document storage device, the retrieval device comprising: logical structure analysis means for analyzing the logical hierarchical format of the documents and for obtaining structural elements and a hierarchical relationship between the structural elements within each document; classification unit designation means for designating a classification unit showing which level the structural elements to be classified are at within the hierarchical relationship; fundamental vector generation means for extracting a key word from the content of each structural element of the classification unit that is designated by the classification unit designation means, and for generating a fundamental vector based on the extracted key word; content vector generation means for extracting a key word from the content of each structural element other than the structural element used to generate the fundamental vector, and for generating a content vector for each structural element based on the extracted key word; vector synthesis means for generating, for each structural element of the classification unit, a composite vector based on the corresponding fundamental and content vectors; and classification means for calculating a degree of similarity among the composite vectors and for classifying the structural elements of the documents based on the degree of similarity.

Claim 11

The document retrieval device of claim 7, wherein the content vector generation means provides weights to the content vector based on whether the structural element includes headings.

Claim 12

The document retrieval device of claim 7, wherein: the content vector generation means generates an element vector relative to each structural element other than the structural element used to generate the fundamental vector, calculates the degree of similarity between each element vector and each fundamental vector, and wherein the content vectors are generated based on the generation and weight provided at the element vectors according to the degree of similarity.

Claim 13

The document retrieval device of claim 7, further comprising: classification result maintenance means for maintaining the classification results; and reclassification designation means for reclassifying the structural elements based on the results maintained in said classification result maintenance means.

Claim 14

The document retrieval device of claim 7, further comprising: a display element selection section that selects the structural element of the classification unit from the result of the classification; a display range selection section that allows selection of a range of the structural element to be displayed from the structural elements of the hierarchical structure; and a display element extraction section that extracts from the document storage device the structural element selected by the display element selection section and that displays the range of the structural element on a display means.

Claim 15

The document retrieval device of claim 11, wherein said content vector generation means provides weights to the content vector based on whether the structural element includes headings.

Claim 16

A document retrieval device that classifies a group of documents, each document having structural elements arranged in a logical hierarchical format, each structural element having a content, the documents stored in a document storage device, the retrieval device comprising: logical structure analysis means for analyzing the logical hierarchical format of the documents and for obtaining structural elements and a hierarchical relationship between the structural elements within each document; retrieval unit designation means for designating a retrieval unit, the retrieval unit showing which level the structural elements to be retrieved are at within the hierarchical relationship; fundamental vector generation means for extracting a key word from the content of each structural element of the retrieval unit designated by the retrieval unit designation means, and for generating a fundamental vector based on the extracted key word; content vector generation means for extracting a key word from the content of each structural element other than the structural element used to generate the fundamental vector, and for generating a content vector for each structural element based on the extracted key word; vector synthesis means for generating, for each structural element of the retrieval unit, a composite vector based on the corresponding fundamental and content vectors; query input means for inputting a query; query vector generation means for generating a query vector from the query; and retrieval means for calculating a degree of similarity between the query vector and each composite vector and for retrieving the structural elements of the retrieval unit having a predetermined degree of similarity relative to composite vectors.

Claim 17

A document retrieval device that classifies a group of documents, each document having structural elements arranged in a logical hierarchical format, each structural element having a content, the documents stored in a document storage device, the retrieval device comprising: logical structure analysis means for analyzing the logical hierarchical format of the documents and for obtaining structural elements and a hierarchical relationship between the structural elements within each document; retrieval unit designation means for designating a retrieval unit, the retrieval unit showing which level the structural elements to be retrieved are at within the hierarchical relationship; fundamental vector generation means for extracting a key word from the content of each structural element of the retrieval unit designated by the retrieval unit designation means, and for generating a fundamental vector based on the extracted key word; content vector generation means for extracting a key word from the content of each structural element other than the structural element used to generate the fundamental vector, and for generating a content vector for each structural element based on the extracted key word; vector synthesis means for generating, for each structural element of the retrieval unit, a composite vector based on the corresponding fundamental and content vectors; query input means for inputting a query; query vector generation means for generating a query vector from the query; and retrieval means for calculating a degree of similarity between the query vector and each composite vector and for retrieving the structural elements of the retrieval unit having a predetermined degree of similarity relative to composite vectors.

Claim 18

The document retrieval device of claim 13, wherein the query is a key word list or a natural language text.

Claim 19

The document retrieval device of claim 13, wherein the vector synthesis means provides weights to the content vectors according to the degree of similarity between the fundamental vectors and the content vectors.

Claim 20

The document retrieval device of claim 13, further comprising: a display element selection section that selects the structural element to be displayed from the results of the retrieval; a display range selection section that allows selection of a range of the structural element to be displayed from the retrieved upper structural elements of the hierarchical structure; and a display element extraction section that extracts from the document storage device the structural element that were selected by the display range selection section and that displays the structural elements on a display means.