Automatic identification of Hindi complex predicates for computer-aided reading tool

Satenik Mkhitaryan (1)

1 ERTIM - Équipe de Recherche en Textes, Informatique, Multilinguisme

This study demonstrates how Natural Language Processing resources can be deployed in Computer Assisted Language Learning for Hindi as foreign language learners. Many researchers have shown the importance of bringing attention and awareness on language categories and forms in second language learning. We introduce a web based implementation that provides visual enhancement of texts in order to make language learning targets more salient for the learner. Learners get to choose a text they want to read and the system displays an enhanced version of the text. It supports visual enhancement for some simple categories (nouns, adjectives, verbs etc.) and provides the lemma if asked, using a Hindi Part-Of-Speech tagger. It also detects the complex predicates (CP). Hence, in this study we will focus on automatic identification of Hindi CPs which are known to be problematic for Hindi learners. We assume that highlighting the CPs will help the reader to grasp them as a single unit of verb and not each element of the CP separately, thus contributing to its comprehension and facilitating its acquisition. We will first review the existing methods for detecting Hindi complex predicates which are often qualified as « pain in the neck for NLP ». We will then present a short overview of the tools and resources for processing Hindi. Finally, we will describe our method for detecting CPs in the context of the abovementioned reading tool and we will discuss the results. Motivation

Mots clés: Hindi Complex Predicates; CALL; reading tool; elearning

Domaines: Environnements Informatiques pour l'Apprentissage Humain

Satenik Mkhitaryan. Automatic identification of Hindi complex predicates for computer-aided reading tool. International Conference on Hindi Studies 2016, Ghanshyam Sharma, Sep 2016, Paris, France. ⟨hal-01381745⟩