Description
Building a chat-bot with deep learning
Building a chat-bot with deep learning – Abstract: Most of the time, students must visit universities or web pages to collect various information e.g. semester schedule, grade, curriculum, advisor info, etc. This process is time-consuming and requires manpower to provide such required information. Nowadays, social media and online services such as email, direct message, or live chat are becoming popular. Many students use live chat service because it is easy, fast, and comfortable. However, there are some issues that occurred an extensive amount of valuable time to answer by admin is wasted. Hence, in order to solve the problem properly, this article proposes an approach for building a question-answering system via chatbot technology. Because chatbots are enabling human-machine interactions with the system by the human language, which is very intuitive and user-friendly. This research objective was to develop a question-answer system in Thai using chatbot technology, which was integrated with the LINE mobile application. The experiment added a keyword in the educational domain to the dictionary and compared word segmentation for 3 methods with text classification, a combination of natural language processing (NLP) and Deep Neural Network (DNN) is used in the form of intent for text classification in this article.
Introduction:-
Communication, transport, business, education, public health, industry, etc. have developed continuous improvement. With support from both developments hardware and software computer. Software computers support both manufacturers and customers in chatbots. Nowadays, the most popular chat agent is a chatbot because it is user-friendly enables interactions with the human-machine by the human language. chatbot hides its complexity of information behind. Chatbots hide the complexity of information behind them trying to look like conversations similar between people. Chatbot development is based on 2 principles, that are knowledge base and Machine learning. Example, a Rule-oriented chatbot based on a knowledge base such as ELIZA [1] or ALICE [2]. Data-oriented chatbots based on learning models from samples of dialogues using sophisticated algorithms by a natural language processing (NLP) and machine learning approach.
The complication of chatbots is computed methods to understand and process the Thai language (Thai NLP) is a vital part of processing the text. In practice, we have to add a keyword into the dictionary and compare word segmentation for 3 methods that is NEWMM [3] Cutkum [4], and Deepcut [5]. The conversation chats in Thai using chatbot technology, which was integrated with Dialogflow[6] and Firebase google technology with LINE Messaging API. Combination of Thai natural language processing (Thai NLP) and Deep Neural Network (DNN) used in the form of intent for text classification. The chatbots were developed for supporting education in the Computer Science Department in the School of Information and Communication Technology at the University of Phayao. The name of the chatbot is Latin. It can provide three main users in the education domain including students, teachers, and administrators. The topic is subjects, academic calendar, schedule, grade, and grade point average. The notification and assign exercises or homework and manage the knowledge base. The remaining of this paper is organized as follows. Section II explains related work. Section III presents the theories and techniques used in this paper. Section IV describes details of our proposed approach. Section V discusses the evaluation results. Finally, conclusions are presented in Section VI.
The existing model of the system:-
Facilities that provide customers with inquiries such as call center, web page, or direct message. Tools for supporting in the context of e-banking, e-commerce, and e-education are chatbots. A domain online course such as learning foreign languages and teaching the specific courses. Chatbots are like a helper and sometimes that will keep your students from being boring. Answer or response from chatbots is answering users’ questions within a knowledge base.[7,8]. AI-Base Chatbot applied With the Self-learning on “Intelligent Tutoring Systems (ITS)” at the National Academy of Engineering Grand Challenges, it has the potential to enhance student learning [9]. Chatbot developed by rule-based or rule-oriented chatbots, It is processing for understanding messages and responding to users is dependent on the size and quality of knowledge base or database. On large domains, there are difficulties in controlling the quality and require a long time to be created the knowledge bases. Chatbot developed by data-oriented or Al-baes chatbot. It is processing for understanding messages using learning models from the dataset of dialogues for the training model. Al-baes chatbot has two models that are the difference, that is model for predefined and learns using the repository to the response. Other than based on sequence-to sequence [6] using Machine Translation [10] techniques.
https://youtu.be/wAKHGhaqSvA
Proposed model of the system:-
chat-bot with deep learning In this system, the chatbot provides three main users in the education domain including students, teachers, and administrators. The chatbot provides students that are studies, subjects, academic calendar, schedule, grade, and grade point average. Teachers can notify and assign exercises or homework in chat and the administrator manages the knowledge base. The chatbot system was developed by Dialogflow and Firebase google technology with LINE Messaging API. We implement neural networks for our input data.
System architecture:-
System requirements:-
Hardware requirements:-
- The operating system of Windows 7,8,10(32-bit or 64-bit )
- RAM-4GB
Software requirements:-
- Anaconda navigator
- jupyter notebook
Conclusion:-
chat-bot with deep learning – This paper focuses on Thai chatbots in the education domain using natural language processing (NLP) and Deep Neural networks (DNN). NLP is used in the pre-processing the Thai sentence. The research was experimented with three methods of words segmentation in Thai. Deep Neural Network Model with text classification, prepared the intent dataset for model training and developing Latin Chatbot in Thai with LINE Applications
Reference:_
[1] J. Weizenbaum, “ELIZA – A Computer Program For the Study of Natural Language Communication Between Man And Machine”, Communications of the ACM, vol 9, 1966.
[2] B. AbuShawar, and E. Atwell, “ALICE Chatbot: Trials and Outputs”, Computaciony Sistemas, Vol. 19, No. 4, pp. 625- 632, 2015.
[3] W. Phatthiyaphaibun, K. Chaovavanich, PyThaiNLP [Online]. Available:https://github.com/PyThaiNLP/pythainlp.
[4] P. Boonkwan, and T. Supnithi, “Bidirectional Deep Learning of Context Representation for Joint Word Segmentation and POS Tagging”, International Conference on Computer Science, Applied Mathematics and Applications(ICCSAMA), 2017, pp. 184-196.
[5] R. Kittinaradorn, “DeepCut: A Thai word tokenization library using Deep Neural Network”, Sep 23 2019. [Online]. Available: https://pypi.org/project/deepcut/.
[6] I. Sutskever, O. Vinyals, and V. L. Quoc, “Sequence to Sequence Learning with Neural Networks”, Advances in Neural Information Processing Systems,2014.
[7] B. Heller, and M. Proctor, “Freudbot: An investigation of chatbot technology in distance education”. Proceedings of Educational Media and Technology, Montreal, Canada, pp. 3913-3918, 2005.
[8] A. Kerly, and P. Hall, “Bringing chatbots into education: Towards natural language negotiation of open learner models”, KnowledgeBased Systems, pp.177-185, 2007.
[9] P. Bii, “Chatbot technology: A possible means of unlocking student potential to learn how to learn”. Educational Research, University of Kabianga, 218-221, 2013.
Customer Reviews
There are no reviews yet.