Recent advances in Natural Language Processing, along with easier access to very powerful Language Models and libraries to manipulate them have led to a notable expansion of the NLP application field. From this evolution arise new research issues. If it is not straightforward to find new tasks, per se, traditional tasks such as classification, extraction, summarization or translation are pushed towards handling more complex phenomena, from heterogeneous textual data, possibly in combination with new modalities.
New applicative domains have emerged these past few years, paving the way for NLP use on very diverse data. A community has for example structured itself around NLP for Finance and Economy. E-commerce, security, the legal sector, public administrations (workshops FNP-FNS 2020, ECNLP 3, LT4GOV, etc.) and more recently music (NLP4Musa) are also subject to dedicated work.
Other domains that have already been covered for several years, such as journalism, social media, digital humanities or the bio-medical domain, are facing an expansion of their field of research. They are now addressed from new perspectives, revealing a sign of an outstanding dynamism of activities in these domains. For instance, beyond the generalist workshop “NLP meets journalism” created in 2015, there are now workshops dedicated to exploiting Knowledge Graphs for journalism or to socio-politic events extraction. Social media analysis or more generally “noisy user-generated texts” analysis is addressed from various angles such as censorship, propaganda and disinformation, threats in online conversations, opinion modeling, fact checking or harassment (workshops WNUT, NLP4IF, PEOPLE 2020, FEVER, etc.). As for the bio-medical domain, the number of dedicated workshops has risen around ten for the year 2020.
The above-mentioned list of applicative domains is not exhaustive. These new domains, beyond the scientific issues related to the specificity of data, also raise numerous ethical questions regarding the use of underlying technologies, the exploitation of automatic processing results, and the new practices that can emerge in the corresponding professional areas. At last, the fact that NLP applications are now largely in contact with the general public raises issues linked to production constraints such as efficiency, cost, robustness, and resilience to adversarial attacks. We are interested in this special issue of TAL in all these scientific, technological and ethical questions, yielded by the expansion of NLP within a variety of application domains.
For this thematic issue, the TAL journal largely calls for contributions related to these “new” application domains, as described above, or to “new” paradigms in already covered domains, in particular, but not limited to:
- scientific issues related to the specificities of data,
- implemented solutions to address new application domains,
- new paradigms in less recent domains,
- methodological issues,
- ethical and social implications of research in these new domains.