From opinion mining to document search: an
overview of applicative use-cases along with practical
implementation issues and open research questions
Géraldine Damnati1,†
1
    Orange Innovation, Lannion, France


                                         Abstract
                                         Recent advances in Natural Language Processing and Information Retrieval, 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. Applicative domains that were traditionally exploiting language technologies like
                                         bio-medicine, journalism, social media, customer relationship or digital humanities are progressively
                                         addressing more complex tasks, from more complex and heterogeneous data. In the meantime, research
                                         communities are structuring themselves around new applicative domains like finance, e-commerce,
                                         justice, legal or administrative organizations. From an operational point of view, the prevalence of Data
                                         Scientist positions is growing in many companies and public institutions and language technologies
                                         skills has become key in many applicative domains. However, despite the huge acceleration of progress
                                         in NLP and IR in the deep learning era, state-of-the-art models still represent practical implementation
                                         and cost issues in order for them to be deployed in operational services. Additionally, the mismatch
                                         between public datasets and business data remains a limitation to bridge the gap between the research
                                         community and operational applications.
                                             In this talk I will present several applicative use-cases developed at Orange Innovation in order
                                         to feed operational services towards customers and employees. I will discuss the current bottlenecks
                                         for the deployment of state-of-the art models in operational services. From a broader perspective, I
                                         will present how tasks that are studied in research can yield new opportunities and, conversely, how
                                         feedbacks from operational projects can yield new research directions. For instance, as an international
                                         group, multilinguality plays a central part in our research activities. I will also highlight how we manage
                                         to bridge the gap between academic and collaborative research and business use cases through the
                                         construction of specific datasets that can be of interest for the community and transpose to operational
                                         contexts.

                                         Keywords
                                         Information Retrieval, Natural Language Processing, Information Extraction, Opinion mining, Techno-
                                         logical transfer




CIRCLE (Joint Conference of the Information Retrieval Communities in Europe), July 04–07, 2022, Samatan, Gers, France
$ geraldine.damnati@orange.com (G. Damnati)
 0000-0003-2218-4181 (G. Damnati)
                                       © 2022 Copyright 2022 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
    CEUR
    Workshop
    Proceedings
                  http://ceur-ws.org
                  ISSN 1613-0073
                                       CEUR Workshop Proceedings (CEUR-WS.org)