Association Measures

Association measures quantify the degree to which two terms or concepts are associated based on their co-occurrence found in text.

UMLS-Association -- is a freely available software package that can be used to quantify the strength of association between UMLS concepts using co-occurrence information from the MetaMapped Medline Baseline. Given two medical terms as input, the output will be a numeric score that indicates how associated the terms are. It utilizes the Ngram Statistics Package which implements a variety of standard measures association including Fisher's exact test, the log likelihood ratio, Pearson's chi-squared test, the Dice Coefficient, etc.

Indirect association and ranking hypotheses for literature based discovery. Sam Henry and Bridget T. McInnes. BMC bioinformatics 20 (1), 425, 2019.
Association measures for estimating semantic similarity and relatedness between biomedical concepts. Sam Henry, Alex McQuilkin and Bridget T. McInnes. Artificial intelligence in medicine 93, 1-10, 2019.
Linking Term Association: ranking implicit terms in literature-based discovery. Sam Henry. In Proceedings of the American Medical Informatics Association Graduate Student NLP-WG Workshop. November 2017. (poster)
Resolving Structural Ambiguity of Medical Terms with Statistical Model Fitting. Serguei V. Pakhomov and Bridget T. McInnes, Linguistic Society of America (LSA) Presentation, 2005. (code: NSP)
Extending the Log-Likelihood Ratio to Improve Collocation Identification. Bridget T. McInnes, Master of Science Thesis, Department of Computer Science, University of Minnesota, Duluth, December, 2004. (code: NSP)
Incorporating Bigram Statistics to Spelling Correction Tools. Bridget T. McInnes, Serguei V. Pakhomov, Ted Pedersen and Christopher G. Chute. Poster Presentation, MEDINFO 2004. (code: NSP)

ASSOCIATION MEASURES