A research on text classifiers to build a college admission chatbot
Abstract
In text classification field, most of previous studies only measured and evaluated the selected model upon one test set with a certain size, as well as did not clarify the training time of each model. This paper focuses on comparing and evaluating 3 text classification models: Support Vector Machine, Linear Regression using SGD, Naïve Bayes in terms of accuracy with different test sets, then clarifying the evaluation parameters with a test set of 900 input questions. Besides, the reseach compares the training time of each model at different training sets with varied training-sizes. The results have shown that Naïve Bayes has good accuracy and the training time is also significantly dominant when compared with the two others. Afterall, the author uses the above research results to propose a solution to build an admission chatbot through Facebook, provide promising empirical results and make it applicable to educational units in Vietnam.
Moraref