With this VM grant we aim to merge the existing classification methods within Natural Language Processing (NLP) to detect and classify both communicative intentions and sentiment analysis in a Spanish corpus with the purpose of verifying if the identification of the latter improves the automatic detection of the former. Very interesting results have already been shown for this task in the context of English social media corpora, as it has been demonstrated that sentiment analysis helps to a more accurate automatic detection of communicative intentions indeed. By addressing this intention and sentiment challenge, the main idea is to explore ways of integrating this semantic-pragmatic information into a Natural Language Generation (NLG) system which is capable of including those intention and sentiment linguistic patterns in its architecture so text can be automatically generated according to a determined intention and sentiment. In this way, we aim to improve the general common sense of the architecture as well as to take into consideration more semantic-pragmatic aspects of language in NLG systems. We want to also contribute to the research community by sending the results from this research collaboration to any forthcoming international NLP conference to motivate scientific dissemination.
Multi-task, Multilingual, Multi-modal Language Generation
