Back to results
US 7,584,189 B2Granted
Sentence classification device and method
AssigneeYamatake Corporation
InventorsEiji Murakami, Motomi Kohata, Takao Terano
Life of this patent
Estimating the term…
Abstract
A sentence classification device includes a storage unit, arithmetic processing unit, and screen display unit. The storage unit stores a document set of a plurality of documents and a term list including a plurality of terms each having one or more words. The arithmetic processing unit classifies the respective documents stored in the storage unit on the basis of the term list. The screen display unit displays a processing result obtained by the arithmetic processing unit on a screen.
Claims9
A sentence classification device comprising:
- a storage unit which stores various kinds of processing information including a document set of a plurality of documents and a term list including a plurality of terms each having not less than one word;
- an arithmetic processing unit which classifies the respective documents stored in said storage unit on the basis of the term list;
- and a screen display unit which displays a processing result obtained by said arithmetic processing unit on a screen,
- said arithmetic processing unit comprising a DT matrix generation unit which generates a DT matrix in which a relationship between each document stored in said storage unit and each term is two-dimensionally expressed,
- a DT matrix transformation unit which generates a transformed DT matrix in which documents are divided into blocks for each cluster by transforming the DT matrix on the basis of a DM decomposition method used in a graph theory,
- a virtual representative generation unit which generates a virtual representative document which virtually represents each cluster on the transformed DT matrix from terms of documents belonging to each cluster,
- a large classification generation unit which generates large classifications of documents by repeating clustering processing as a step by using the DT matrix generated by said DT matrix generation unit in an initial state,
- the clustering processing comprising a process of causing said DT matrix transformation unit to generate a transformed DT matrix from the DT matrix,
- a process of causing said virtual representative generation unit to generate a virtual representative document for each cluster on the generated transformed DT matrix,
- a process of generating a new DT matrix used in next clustering processing by adding the generated virtual representative documents to the transformed DT matrix and deleting documents belonging to the clusters corresponding to the virtual representative documents from the transformed DT matrix, and
- a process of storing at least information associated with documents constituting each cluster as large classification data in said storage unit for each cluster, and
- a display processing unit which causes said screen display unit to display, on a screen, large classifications of documents in a tree structure using cluster symbols representing clusters generated in clustering processing and link symbols indicating integral relations between the clusters on the basis of the large classification data stored in said storage unit.
9 more elements
A sentence classification method comprising the steps of:
- generating a DT matrix in which a relationship between each document and each term is two-dimensionally expressed by referring to a storage unit which stores various kinds of processing information including a document set of a plurality of documents and a term list including a plurality of terms each having not less than one word;
- generating a transformed DT matrix in which documents are divided into blocks for each cluster by transforming the DT matrix on the basis of a DM decomposition method used in a graph theory, generating a virtual representative document which virtually represents each cluster on the generated transformed DT matrix from terms of documents belonging to each cluster;
- generating a new DT matrix by adding the generated virtual representative documents to the transformed DT matrix and deleting documents belonging to the clusters corresponding to the virtual representative documents from the transformed DT matrix;
- storing at least information associated with documents constituting each cluster as large classification data in the storage unit for each cluster;
- generating large classifications of documents by repeatedly performing clustering processing comprising the steps of generating a transformed DT matrix by using a new DT matrix, further generating a new DT matrix by generating virtual representative documents, and storing information;
- and
- displaying, on a screen, large classifications of documents in a tree structure using cluster symbols representing clusters generated in clustering processing and link symbols indicating integral relations between the clusters on the basis of the large classification data stored in the storage unit.
4 more elements
A recording medium on which a program is recorded, the program causing a computer to execute the steps of:
- generating a DT matrix in which a relationship between each document and each term is two-dimensionally expressed by referring to a storage unit which stores various kinds of processing information including a document set of a plurality of documents and a term list including a plurality of terms each having not less than one word;
- generating a transformed DT matrix in which documents are divided into blocks for each cluster by transforming the DT matrix on the basis of a DM decomposition method used in a graph theory, generating a virtual representative document which virtually represents each cluster on the generated transformed DT matrix from terms of documents belonging to each cluster;
- generating a new DT matrix by adding the generated virtual representative documents to the transformed DT matrix and deleting documents belonging to the clusters corresponding to the virtual representative documents from the transformed DT matrix;
- storing at least information associated with documents constituting each cluster as large classification data in the storage unit for each cluster;
- generating large classifications of documents by repeatedly performing clustering processing comprising the steps of generating a transformed DT matrix by using a new DT matrix, further generating a new DT matrix by generating virtual representative documents, and storing information;
- and displaying, on a screen, large classifications of documents in a tree structure using cluster symbols representing clusters generated in clustering processing and link symbols indicating integral relations between the clusters on the basis of the large classification data stored in the storage unit.
3 more elements
Dependencies are inferred from the claim text.