Comparative analysis of chatbot performance using different word embedding techniques for student query response

College

College of Computer Studies

Department/Unit

Software Technology

Document Type

Conference Proceeding

Source Title

2025 International Workshop on Artificial Intelligence and Education

Publication Date

2025

Abstract

A chatbot can be developed to respond to student queries. Given the variety of chatbot implementations utilizing different word embedding techniques, it is essential to evaluate their performance in this context. This study aimed to analyze and compare the effectiveness of various chatbot models employing different word embedding methods in answering student queries. Three chatbots were created where each chatbot has a different word embedding technique, namely Bag-of-Words (BoW), Term Frequency-Inverse Document Frequency (TF-IDF), and Word2Vec. A feedforward neural network was employed across all implementations to classify the appropriate responses. The University ChatBot Queries and Responses dataset from Kaggle was used to simulate real-world student queries. The chatbot models were evaluated using four metrics: accuracy, F1-score, word embedding time, and training time. Among the three models, the TF-IDF-based chatbot demonstrated the best overall performance. However, the study was conducted using a relatively small dataset, which may limit the generalizability of the results. Therefore, future research is recommended to explore these implementations using larger and more diverse datasets to validate and extend the findings.

html

Disciplines

Software Engineering

Keywords

Chatbots

Upload File

wf_no

This document is currently not available here.

Share

COinS