Related Experiment Videos
Methods for semi-automated indexing for high precision information retrieval.
Daniel C Berrios1, Russell J Cucina, Lawrence M Fagan
1Stanford Medical Informatics, Stanford University, Stanford, California 94035, USA. berrios@email.arc.nasa.gov
Journal of the American Medical Informatics Association : JAMIA
|October 19, 2002
Summary
The Internet-based Semi-automated Indexing of Documents (ISAID) system significantly speeds up the creation of detailed and accurate textbook indexes compared to manual methods. Natural language processing enhances indexing speed and precision for users.
Area of Science:
- Information Science
- Medical Informatics
- Document Indexing
Background:
- Traditional manual indexing of documents can be time-consuming and may not always yield the desired level of detail or accuracy.
- Developing efficient and effective indexing systems is crucial for organizing and accessing large volumes of information, particularly in academic and medical fields.
Purpose of the Study:
- To evaluate the efficacy of a new system, ISAID (Internet-based Semi-automated Indexing of Documents), for generating textbook indexes.
- To determine if ISAID can produce indexes that are more detailed and useful to readers compared to manual indexing.
Main Methods:
- A pilot study involved a nonrandomized trial comparing ISAID with manual indexing.
- A methods evaluation used a randomized, cross-over trial comparing three versions of ISAID with 12 physicians over 36 indexing sessions.
- Measurements included index term tuples per minute (TPM), inter-indexer consistency, and system usability ratings.
Main Results:
- ISAID significantly decreased indexing times compared to manual methods.
- Inter-indexer consistency with ISAID ranged from 15% to 65% across different documents.
- Users of the full ISAID version were faster, generating an average of 5.6 TPM, with significant learning effects observed, reaching 9.1 TPM by the third session.
Conclusions:
- The ISAID system enables users to create complex, precise, and accurate indexes for full-text documents substantially faster than manual methods.
- The natural language processing capabilities of ISAID contribute significantly to improved indexing speed and accuracy.