Predictive Modeling: Guessing the NLP Terms of Tomorrow

Predictive modeling: guessing the NLP terms of tomorrow Gil Francopoulo 1, Joseph Mariani 2, Patrick Paroubek 2 1 LIMSI, CNRS, Université Paris-Saclay + Tagmatica (France) 2 LIMSI, CNRS, Université Paris-Saclay (France) gil.francopoulo@wanadoo.fr, joseph.mariani@limsi.fr, pap@limsi.fr Abstract Predictive modeling, often called “predictive analytics” in a commercial context, encompasses a variety of statistical techniques that analyze historical and present facts to make predictions about unknown events. Often the unknown events are in the future, but prediction can be applied to any type of unknown whether it be in the past or future. In our case, we present some experiments applying predictive modeling to the usage of technical terms within the NLP domain. Keywords: Predictive Modeling, Predictive Analytics, Term Extraction, Natural Language Processing 1. Introduction Predictive modeling, often called “predictive analytics” in a commercial context, encompasses a variety of statistical techniques that analyze historical and present facts to make predictions about unknown events. Often the unknown events are in the future, but prediction can be applied to any type of unknown whether it be in the past or future. In our case, we present some experiments applying predictive modeling to the usage of technical terms within the NLP domain. 2. Context Our work comes after the various studies initiated in the Workshop entitled: “Rediscovering 50 Years of Discoveries in Natural Language Processing” on the occasion of ACL’s 50th anniversary in 2012 [Radev et al 2013] where a group of researchers studied the content of the corpus recorded in the ACL Anthology [Bird et al 2008]. Different studies were presented from reuse detection [Gupta et al 2012] to topic detection [Anderson et al 2012]. 3. Corpus Our research began by gathering a large corpus of NLP scientific articles covering documents produced from 1965 up to 2015. This corpus gathers a large content of our own research field, i.e. NLP, covering both written and spoken sub-domains and extended to a limited number of corpora, for which Information Retrieval and NLP activities intersect. This corpus was collected at LIMSI-CNRS (France) and is named NLP4NLP [Francopoulo et al 2015]. It contains currently 65,003 documents coming from various conferences and journals with either public or restricted access. This represents a large part of the existing published articles in our field, aside from the workshop proceedings and the published books. It should be noted, that most other studies in our domain are based on the ACL Anthology1 which is dedicated to text processing, but as LIMSI and IMMI laboratories work both in the written and spoken language processing domains, we chose to address both. The ACL Anthology (although extremely valuable) is 1 http://aclweb.org/anthology 2 http://www.isca-speech.org/iscaweb 3 http://www.signalprocessingsociety.org only approximately one third of our corpus, the majority of the other papers coming from the ISCA 2 and IEEE 3 archives. The ACL Anthology and ISCA archives are in open access, see the details of the 34 sub-corpora in table 1. Let’s note that for a joint conference (which is a rather infrequent situation), the paper is counted once in each row …

Content truncated to 500 words