ChatbotDViL: Diseño y implementación de una interfaz en lenguaje natural para visualización de datos

Abstract

[en] Nowadays there are different tools for natural language processing to, for example, perform emotion analysis in messages sent in social networks or to detect hate speech. Within this conceptual framework, the present work continues previous projects in which a group of linguists analyzed conversations from different news articles, labeling each of the comments with different characteristics (constructiveness, sarcasm, level of toxicity, etc.), and elaborating different hypotheses regarding the toxicity of comments and hate speech. In this work, a platform called DViL (Data Visualization in Linguistics) was created to process and represent news comments by means of graphs. In this way, visualization in the form of a graph can help linguists to better analyze their labeling of messages. In this work we propose a series of improvements to the DViL platform, although the main focus of this work is on the design and development of a chatbot, thanks to which a user can make natural language queries to analyze data visualizations, either by speaking or typing. This chatbot is able to understand the user’s intentions thanks to a system based on ML (Machine Learning) that, through the training of a set of examples, is able to classify the user’s intention and find the best answer to the user’s request. This response can be a text and/or a graph with basic statistics, as well as an update of the main visualization being displayed by the platform or a new visualization in a popup window (subgraph).