[Fad17a] Feedback : use of deep learning in the context of Poorly endowed languages.

Conférence Nationale avec comité de lecture : Traitement Automatique des Langues Naturelles (TALN), June 2017, pp.10, France,

Auteurs: H. Fadili

Mots clés: Poorly endowed languages, deep learning, language model, semantics, word embeddings

Résumé: Feedback : use of deep learning in the context of Poorly endowed languages. It is estimated that there are several thousand languages spoken in the world and only a few dozen have resources (tools, corpuses, annotations, etc.) for automatic processing. Those with little or no resources are called poorly endowed languages (PEL). Several UNESCO reports state that most poorly endowed languages are endangered. In addition, several language specialists believe that their disappearance is accelerated by the phenomena of new technologies (Internet, social networks, etc.) which further marginalize them. However, according to the same specialists, the integration of poorly endowed languages into the world of new technologies could constitute an opportunity for their development, their preservation and therefore for their survival. Indeed, make available to the users, tools encouraging them to discover and to create in the PEL, helped by bridges with better endowed languages (BEL), such as functionalities of: links, alignment, Translation, analysis, etc. Could have a positive impact on the popularity of their use and consequently on their development. In this article, we present an experiment exploiting the new technologies of deep learning in the context of the semantic analysis of PEL. The aim is to show through an example of approach that we can exploit certain technologies that are easily adaptable to languages suffering from the lack of resources in terms of content and computer tools; Hoping that it will also help to raise awareness and encourage researchers in the field to propose generic solutions integrating in their design the support of LPD.


@inproceedings {
title="{Feedback : use of deep learning in the context of Poorly endowed languages.}",
author=" H. Fadili ",
booktitle="{Traitement Automatique des Langues Naturelles (TALN)}",
address=" France",