TY - JOUR AU - Sangkrajang, Kanita AU - Tangwannawit, Panana PY - 2020/09/14 Y2 - 2024/03/28 TI - MESSENGER’S CHATBOT BASED ON ARTIFICIAL INTELLIGENCE FOR DIGITAL LIBRARY SERVICE JF - Life Sciences and Environment Journal JA - Life Sci. Environ. J. VL - 21 IS - 2 SE - Research Articles DO - UR - https://ph01.tci-thaijo.org/index.php/psru/article/view/242317 SP - 468-480 AB - <p>This research aims to develop the messenger’s chatbot based on artificial intelligence for digital library service. This machine learning application is a branch of artificial intelligence, using a supervised learning method to learn the conversations of user in large quantities of datasets. Data classification techniques were used classify word or sentence attributes with models that calculate the probability of a character set and then process the Natural Language Processing (NLP). Then analyzed by the Winnow algorithm, the model predicts in each sentence what the user wants to communicate and how to respond. These research tools consisted of AngularJS (Front-End), NodeJS (Back-End), MongoDB (Amazon Web Service), Firebase (Front-End), Heroku (Back-End) and Bitbucket. The development of a Facebook messenger’s chatbot based on artificial intelligence for digital library service was conducted with 3,361 messages as samples of chat data out of 220 total respondents (1,987 messages from users and 1,374 from librarians). All the messages from the users were analyzed and categorized based on their features with feature selection techniques. The results showed that 1,254 messages after analyzed with this technique were assigned Labels into N-Gram Function by using N-Gram of Letter to cut into syllables with N=3. Next, the performance of the model intended to be used with the Facebook messenger’s chatbot for digital library service was tested through Cross Validation K-Fold with K=10. The results of the model were classified into 3 different levels of deep learning. After the classification of the data based on the 3 levels of deep learning, results showed that the average accuracy of the 15 models was at 95.28%. To conclude, the messenger’s chatbot based on AI has sufficient performance and accuracy. The model is then applied to the library system.</p><p>&nbsp;</p> ER -