Mining Medline: Abstracts, Sentences, or Phrases?

Abstract

A growing body of work addresses automated mining for biochemical information from digital repositories of scientific literature such as MEDLINE. Some of this work uses abstracts as the unit of text from which to extract facts. Other work uses sentences for this purpose, while still other work uses phrases. Here, we compare abstracts, sentences, and phrases in MEDLINE using the standard information retrieval performance measures of recall, precision, and effectiveness for the task of mining interactions among biochemical terms based on term cooccurrence. Results show statistically significant differences that can impact the choice of text unit, although no one of these three text units is unambiguously superior to the others.

In Silico Analysis of Golgi Glycosyltransferases: A Case Study on the LARGE-Like Protein Family

“…Many public databases and bioinformatics tools have been developed and are currently available for use (Ding & Berleant, 2002). The primary goal of bioinformaticians is to develop reliable databases and effective analysis tools capable of handling bulk amount of biological data.…”

Science and Technology Text Mining: Electric Power Sources

“…In the latter case, if the two co-occurring words/ phrases are physically positioned very far apart, co-occurrence may have no meaning. A recent study quantifies some of the precision-recall tradeoffs for different units, ranging from phrases to Abstracts [33].…”

Getting to the (c)ore of knowledge: mining biomedical literature

“…For protein name tagging, accuracies as high as around 95% have been reported [67], but care should be given to the test set composition. It is known that for some organisms or some protein subdomains, the nomenclature is fairly rigidly standardized and excellent tagging accuracy can be reached there.…”

“…The effect of accidental co-occurrence could be minimized by requiring frequent corroboration of any pairing. Using a similar co-occurrence approach, Ding et al [67] found that precision and recall traded off when the length of the used text segment was varied. Working with phrases gave generally better precision, while working with entire abstracts gave best recall; sentences scored in between.…”