A system of C++ language programs has been developed for the purpose of finding the closely related documents in Medline and for the purpose of performing machine learning on sets of documents. The system has a number of unique features: 1) It is based on a number of C++ classes and highly modular so that alterations in the system are relatively simple to perform. 2) The system currently processes PubMed data by extracting from the Sybase repositories using a C++ interface to Sybase. However, a change in the interface portion of the system would allow it to be applied to any large database consisting of discrete textual records. 3) Data processed by the system is stored as compressed file structures, etc. These structures are updatable so that new data may be continually added to the system as it becomes available. 4) Documents are compared with each other using a Bayesian form of analysis. 5) The latest work on this system has involved adding the ability to generate themes using an EM algorithm approach. Also recently code has been multithreaded and memory mapping capabilities added to speed up processing. The system described here is now not only being used to process all of MEDLINE for our research purposes, but also to produce the related documents for arbitrary pieces of text by other groups here in the NLM and outside of the NLM. We also are currently involved in a project to use the system for mining email communications for the NLM help desk.
| Wilbur, W John; Kim, Won (2003) The dimensions of indexing. AMIA Annu Symp Proc :714-0 |
