NLP-based chatbot for HAMK
Trifunovic, Dejan (2019)
2019
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-201905027224
https://urn.fi/URN:NBN:fi:amk-201905027224
Tiivistelmä
Artificial intelligence and conversational agents have become increasingly important in recent years due to technological progress. In particular, chatbots are attracting growing interest, especially in companies. Häme University of Applied Sciences is also dealing with the subject of chatbots and examines products of various manufacturers, in order to improve the student services by introducing them. Therefore, this study investigates the suitability of an open source chatbots solution of the manufacturer Rasa Technologies GmbH for the service provision of the university in the areas of Student Services and Admissions Services.
By reading scientific reports, reference books, and the manufacturer's technical documentation, knowledge was acquired which served to understand methodical procedures and to apply them correctly. This included collecting and recording training data for the models, developing and training a chatbot prototype and implementing it into an existing test environment.
The most important finding was that a chatbot's responses reflected the training it had received. The quality of the collected data, as well as its amount, are again the prerequisite for good training.
The investigations showed that Rasa's open source product can be considered as a chatbot solution for frequently asked questions. The installation, training, and implementation in an existing environment was feasible and the answers to the questions worked quite well. If the chatbot should be able to manage longer conversational flows, it needs a correspondingly higher training effort, which grows with the increasing amount of training data.
By reading scientific reports, reference books, and the manufacturer's technical documentation, knowledge was acquired which served to understand methodical procedures and to apply them correctly. This included collecting and recording training data for the models, developing and training a chatbot prototype and implementing it into an existing test environment.
The most important finding was that a chatbot's responses reflected the training it had received. The quality of the collected data, as well as its amount, are again the prerequisite for good training.
The investigations showed that Rasa's open source product can be considered as a chatbot solution for frequently asked questions. The installation, training, and implementation in an existing environment was feasible and the answers to the questions worked quite well. If the chatbot should be able to manage longer conversational flows, it needs a correspondingly higher training effort, which grows with the increasing amount of training data.