USI: a fast and accurate approach for conceptual document annotation.

Area of Science:

• Information Science
• Computer Science
• Biomedical Informatics

Background:

• Semantic approaches and concept-based information retrieval require continuous corpus enrichment through expert-driven entity annotation.
• This manual annotation process is time-consuming and demands specialized domain and ontology knowledge.
• Existing strategies for easing the indexing process leverage document features but often fall short in efficiency and accuracy.

Purpose of the Study:

• To introduce the User-oriented Semantic Indexer (USI), a novel method designed for fast and intuitive semantic indexing.
• To present a solution for suggesting conceptual annotations for new entities by leveraging related, already indexed documents.
• To evaluate the performance of USI against existing methods in terms of annotation quality and speed.

Main Methods:

• USI suggests conceptual annotations for new entities by analyzing related, previously indexed documents.
• The method relies on neighbor documents, eliminating the need for a representative learning set.
• Performance is evaluated using standard metrics and semantic similarity measures, compared against text-specific methods and the MeSH thesaurus.

Main Results:

• USI demonstrates superior performance compared to previous methods in both annotation quality and speed for biomedical papers.
• The system provides a consistent annotation scored with a global criterion, outperforming concept-specific scoring.
• Evaluations confirm the effectiveness of USI in enhancing semantic indexing tasks.

Conclusions:

• The User-oriented Semantic Indexer (USI) offers an efficient and effective solution for semantic indexing without requiring a learning set.
• By utilizing neighbor document information, USI achieves higher quality and speed than existing approaches.
• USI's global scoring criterion ensures consistent and reliable conceptual annotations.