Related Experiment Videos
Ontology-based structured cosine similarity in document summarization: with applications to mobile audio-based
1MIS Department, National Chengchi University, Taipei, Taiwan. yuans@seed.net.tw
Summary
This study introduces structured cosine similarity (SCS) for improved document clustering. The novel method enhances text categorization quality, stability, and efficiency by considering document structure.
Area of Science:
- Data Mining
- Knowledge Discovery
- Natural Language Processing
Background:
- Automated text categorization is crucial for managing large document sets.
- Existing term-based clustering methods often overlook document structure, limiting performance.
- Speech documents present unique challenges due to limited explicit features.
Purpose of the Study:
- To develop a novel document clustering method that incorporates document structure.
- To improve the quality, stability, and efficiency of text categorization.
- To address the specific challenge of clustering speech documents for knowledge management.
Main Methods:
- Introduction of Structured Cosine Similarity (SCS), a novel algorithm.
- Modeling document summarization to capture structural information.
- Applying SCS to cluster speech documents from enterprise oral sharing.
Main Results:
- SCS demonstrates improved performance in document clustering quality.
- The method shows enhanced stability and efficiency compared to traditional approaches.
- Promising results achieved in clustering speech documents for knowledge acquisition.
Conclusions:
- Structured Cosine Similarity (SCS) offers a significant advancement in document clustering.
- Incorporating document structure effectively enhances text categorization, especially for challenging data types like speech.
- The method facilitates audio-based knowledge management and sharing within enterprises.