Analyzing the evolutions of the trends of a scientific domain in order to provide insights on its states and to establish reliable hypotheses about its future is the problem we address here. We have approached the problem by processing both the metadata and the text contents of the domain publications. Ideally, one would like to be able to automatically synthesize all the information present in the documents and their metadata. As members of the NLP community, we have applied the tools developed by our community to publications from our own domain, in what could be termed a “recursive” approach. In a first step, we have assembled a corpus of papers from NLP conferences and journals for both text and speech, covering documents produced from the 60’s up to 2015. Then , we have mined our scientific publication database to draw a picture of our field from quantitative and qualitative results according to a wide range of perspectives: ranging from sub-domains, specific communities, chronology, terminology, conceptual evolution, re-use and plagiarism, trend prediction, novelty detection and many more. We provide here an account of the corpus collection and of its processing with NLP technology, indicating for each aspect which technology was used. We conclude on the benefits brought by such corpus to the actors of the domain and on the conditions to generalize this approach to other scientific domains. Conference Topics Methods and techniques, Citation and co-citation analysis, Scientific fraud and dishonesty, Natural Language Processing
[1]
Patrick Paroubek,et al.
Rediscovering 15 + 2 years of discoveries in language resources and evaluation
,
2016,
Lang. Resour. Evaluation.
[2]
Florian Boudin.
TALN Archives : une archive numérique francophone des articles de recherche en Traitement Automatique de la Langue
,
2013
.
[3]
Patrick Drouin,et al.
Detection of Domain Specific Terminology Using Corpora Comparison
,
2004,
LREC.
[4]
Patrick Paroubek,et al.
Rediscovering 25 years of discoveries in spoken language processing: a preliminary ISCA archive analysis
,
2013,
INTERSPEECH.