Transfer-Based Learning-to-Rank Assessment of Medical Term Technicality

Transfer-Based Learning-to-Rank Assessment of Medical Term Technicality Dhouha Bouamor Leonardo Campillos-Llanos Anne-Laure Ligozat Sophie Rosset Pierre Zweigenbaum LIMSI, CNRS, Université Paris-Saclay, 91405 Orsay, France firstname.lastname@limsi.fr Abstract While measuring the readability of texts has been a long-standing research topic, assessing the technicality of terms has only been addressed more recently and mostly for the English language. In this paper, we train a learning-to-rank model to determine a specialization degree for each term found in a given list. Since no training data for this task exist for French, we train our system with non-lexical features on English data, namely, the Consumer Health Vocabulary, then apply it to French. The features include the likelihood ratio of the term based on specialized and lay language models, and tests for containing morphologically complex words. The evaluation of this approach is conducted on 134 terms from the UMLS Metathesaurus and 868 terms from the Eugloss thesaurus. The Normalized Discounted Cumulative Gain obtained by our system is over 0.8 on both test sets. Besides, thanks to the learning-to-rank approach, adding morphological features to the language model features improves the results on the Eugloss thesaurus. Keywords: Technicality of Medical Terms; Learning to rank; Terminology 1. Introduction The difference between the language used by health care professionals and that used by patients is cited as a source of miscommunication (Elhadad and Sutaria, 2007) or diffi- culty when mining patient forums (Nikfarjam et al., 2015). For example, lay people tend to use idiomatic expressions such as “mal de chien”(Fr) [literally, “dog pain” (En)] to refer to “douleur intense”(Fr) [severe pain, En]. Dictionaries which differentiate familiar terms from spe- cialized terms are valuable for many applications of Nat- ural Language Processing such as sentiment analysis (Ali et al., 2013) and paraphrase acquisition (Elhadad and Su- taria, 2007; Deléger and Zweigenbaum, 2009; van der Plas and Tiedemann, 2010). In the context of an e-learning sys- tem for training medical students, we are creating a vir- tual patient who must answer student questions in a nat- ural way, hence using patient language (Campillos-Llanos et al., 2015; Campillos-Llanos et al., 2016). This is yet an- other application which requires a way to assess which term among a set of equivalent expressions would be more likely to be used by a patient. Nevertheless, such resources remain scarce and thus far, only one, the Consumer Health Vocabulary (CHV, Kesel- man et al. (2007)), aims at estimating how familiar a term is. However, the CHV only covers the English language, and limited attempts have been made to cover other lan- guages such as French and Portuguese. Tapi Nzali et al. (2015) mention the creation of a French CHV, but actually address a different problem: the identification of abbrevi- ations, of spelling errors, and of lay terms associated to (but not necessarily equivalent to) specialized terms, such as morphine (morphin, which they consider as a lay term) – douleur (pain, which they consider as a specialized term). Tenorio and Torres Pisa (2015) describe plans to build a Brazilian …

Content truncated to 500 words