An Evolutive Knowledge Base for “AskBot” Toward Inclusive and Smart Learning-based NLP Techniques

Artificial Intelligence chatbots have shown a growing interest in different domains including e-learning. They support learners by answering their repetitive and massive questions. In this paper, we develop smart learning architecture for an inclusive chatbot handling both text voice messages. Thus, disabled can easily use it. We automatically extract, preprocess, vectorize, construct AskBot's Knowledge Base. The present work evaluates various vectorization techniques with similarity measures to answer learners' proposed handles Wh-Questions starting Wh words Non-Wh-Questions, beginning unpredictable words. Regarding Wh-Questions, neural network model classify intents. Our results show that the model's accuracy F1-Score are equal 99,5%, 97% respectively. With score of 0.6, our findings indicate TF-IDF has performed well, correctly 90% tested Wh-Questions. Concerning No-Wh Questions, soft cosine measure, fasttext successfully answered 72% Non-Wh-Question.