A Study of the Morpho-Semantic Relationship in Medline

Abstract

Morphological analysis as applied to English has generally involved the study of rules for inflections and derivations. Recent work has attempted to derive such rules from automatic analysis of corpora. Here we study similar issues, but in the context of the biological literature. We introduce a new approach which allows us to assign probabilities of the semantic relatedness of pairs of tokens that occur in text in consequence of their relatedness as character strings. Our analysis is based on over 84 million sentences from the MEDLINE database, over 2.3 million token types that occur in MEDLINE, and enables us to identify over 36 million token type pairs which have assigned probabilities of semantic relatedness of at least 0.7 based on their similarity as strings. The quality of these predictions is tested by two different manual evaluations and found to be good.

PubTermVariants: biomedical term variants and their use for PubMed search

“…The Morpho-Semantic Analysis. Finally, we make use of the study of morpho-semantic relationships in Medline (Wilbur and Smith 2013). For a candidate pair of strings the approach assigns a probability that the strings are semantically related.…”

“…We then apply the hypergeometric (HG) test (Larson 1982) to decide whether the observed co-occurrence for a particular pair of terms is above random. For the pairs that pass the HG test, we compute the morpho-semantic similarity score following (Wilbur and Smith 2013) and only retain the pairs that score above the threshold.…”

“…A study of morpho-semantic relationships in Medline (Wilbur and Smith 2013) identifies morphologically related tokens in Medline by using character n-grams as features and then computes the probability that two strings are related based on the context. This approach infers the morphological relatedness of two strings in a way more general than stemming, but based on certain substrings of characters on which they match.…”