Document Classification Using Pre-Extracted Feature Words
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing document classification methods require complex and resource-intensive natural language processing to accurately extract feature phrases from documents, making it difficult to efficiently classify large volumes of digitized documents into categories.
Innovation Solution
A document classification method that extracts first feature words from a document, uses a database of pre-classified cases to find similar documents, and calculates a degree of coincidence to determine the most appropriate category, reducing the need for repeated natural language processing and minimizing computational load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complex natural language processing is used to extract feature phrases from documents, then classification accuracy is improved, but computational load and processing time increase significantly
Solution Approach 1:
The patent applies preliminary action by pre-extracting and storing feature words from a large corpus of documents in a database before actual classification tasks. This allows the system to leverage pre-processed information during classification, reducing the need for repeated complex natural language processing on each new document while maintaining high classification accuracy.
2Measurement precision
If repeated dependency analysis is performed to obtain reliability scores, then feature phrase extraction accuracy is improved, but processing time increases
Solution Approach 1:
The system performs dependency analysis and extracts feature phrases in advance on a large corpus of documents, storing the results in a database. During actual classification, the system queries this pre-processed database rather than performing repeated dependency analysis on each new document, significantly reducing processing time while maintaining extraction accuracy.
Solution Approach 2:
The patent creates a copy of pre-processed feature word information from the database to use during classification tasks. Instead of reperforming the complex dependency analysis on each new document, the system uses copied feature word data from the database, which maintains accuracy while dramatically reducing processing time.
3Productivity
If a database of pre-classified cases is used, then computational effort is reduced, but the system complexity increases
Solution Approach 1:
The system performs preliminary classification and feature extraction on a large corpus of documents, storing the results in a database. During actual use, the system queries this database to find similar documents and their categories, which significantly speeds up classification while the database structure manages the complexity centrally.
Data Source
AI summary
A document classification method includes extracting one or more first feature words from first sentence information included in a first document to be classified, extracting, from a database including a plurality of cases, a predetermined number of first cases similar to the first document in accordance with the one or more first feature words extracted from the first sentence information, the plurality of cases including the predetermined number of first cases and being classified in advance into the at least one category, the plurality of cases each including second sentence information indicating a main point of each of one or more items, the one or more items being classified in advance depending on meaning of the item, extracting one or more second feature words of each of the predetermined number of first cases from the second sentence information of each of the predetermined number of first cases, calculating a degree of coincidence between each of the predetermined number of first cases and the first document based on the one or more second feature words extracted from the second sentence information of the first case, and classifying the first document into a category among the at least one category, the category being a category into which a first case having a highest degree of coincidence among the predetermined number of the first cases is classified.