Castelli, MauroDíaz Herrera, Juan Camilo2021-11-162021-11-162021-11-04http://hdl.handle.net/10362/127803Internship Report presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics, specialization in Business AnalyticsArtificial intelligence is changing the way how businesses are affronting their day-to-day difficulties. Chatbots are the perfect demonstration of how simple tasks and queries such as customer support or sales metrics and reporting could be solved without human intervention. This project introduced a task-oriented chatbot framework for Spanish language in a Point-Of-Sale webpage. We applied Natural Language Processing (NLP) techniques such as NER and evaluated two supervised learning methods: (i) an Artificial Neural Network (ANN) and (ii) a Support Vector Machines (SVM) model to create a contextualized chatbot that classifies the user’s intention in a text conversation, allowing bidirectional human-to-machine communication. These intents could go from simple chitchatting to detailed reports, always providing a natural flow in conversation. The results using an augmented and balanced corpus suggested that ANN model performed statistically better than SVM. Additionally, a real-word scenario with a small-talk survey made to five users gave positive feedback about the quality of predictions. Finally, a software architecture using a PaaS computing service and an API framework was proposed to implement this dialog system in further works.engNatural Language ProcessingName Entity RecognitionNatural Language UnderstandingArtificial Neural NetworkSupport Vector MachineWord EmbeddingsChatbotVirtual Assistantmaster thesis202790380