SHS Web Conf.
Volume 139, 2022The 4th ETLTC International Conference on ICT Integration in Technical Education (ETLTC2022)
|Number of page(s)||5|
|Section||Topics in Computer Science|
|Published online||13 May 2022|
The development of a chatbot using Convolutional Neural Networks
Department of Electrical and Computer Engineering, University of Western Macedonia, 501 00, Kozani, Greece
* Corresponding author: email@example.com
Chatbots are artificial intelligence systems that comprehend the intent, context, and sentiment of the user, interact properly with them leading to an increased development of their creation, the past few years. In this study, Convolutional Neural Networks (CNNs) are applied as the classifier and some specific tools for tokenization are used for the creation of a chatbot. Taking into account that it is difficult to apply any algorithm in text, we use a technique called “Word Embedding”, which converts a text into numbers in order to run text processing. Specifically, “Word2Vec” Word Embedding technique was applied. AlexNet, LeNet5, ResNet and VGGNet CNN architectures were utilised. These architectures were compared for their accuracy, f1 score, training time and execution time. The results obtained highlighted that there were significant differences in the performance of the architectures applied. The most preferable architecture of our study was LeNet-5 having the best accuracy compared to other architectures, the fastest training time, and the least losses whereas it was not very accurate on smaller datasets such as our Test Set. Limitations and directions for future research are also presented.
© The Authors, published by EDP Sciences, 2022
This is an Open Access article distributed under the terms of the Creative Commons Attribution License 4.0, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.