Desarrollo de un asistente virtual empleando técnicas de comprensión de lenguaje natural

The objective of this degree work was to develop a virtual assistant based on natural language understanding (NLU) in the Teaching, Research and Liaison Unit (UDIV) of Infrastructure to provide support to the applications of the Computer Science career at Escuela Superior Politécnica Agropecuaria de Manabí Manuel Félix López (ESPAM MFL). The work was based on the Cross-Industry Standard Process for Data Mining (CRISP-DM). First, the main NLU techniques applied in the development of virtual assistants were compiled and Rasa NLU was chosen as the framework for the development of the virtual assistant. Subsequently, an exploratory data analysis was performed, which served to delimit the functions of the virtual assistant. The training data set was formed from the NLU block structure offered by the Rasa framework, the stories, domain, actions, entities, and intentions were defined. Then, the components to assemble the NLU model were detailed, and the training phases were defined. Finally, interaction simulations were performed to evaluate the model. After several tests, the model achieved a high performance with metrics such as an F1 Score of 0.89, a precision of 0.91 and an accuracy of 0.88 in the classification of intentions and entities. Finally, an Application Programming Interface (API) was implemented that marked the integration of the virtual assistant services with a conversational web interface.

• Հիմնական հեղինակ: Cajape Bravo, Jesús Stefano (author)
• Այլ հեղինակներ: Palau Delgado, Sandro Antonio (author)
• Ձևաչափ: bachelorThesis
• Լեզու: spa
• Հրապարակվել է: 2023
• Խորագրեր: Chatbot; Framework; Procesamiento; Acciones; Bloque NLU; Intenciones; APIs
• Առցանց հասանելիություն: http://repositorio.espam.edu.ec/handle/42000/2235