Automatic Chatbot Knowledge Acquisition from Online Forum via Rough Set and Ensemble Learning

Existing chatbot knowledge bases are mostly hand-constructed, which is time consuming and difficult to adapt to new domains. Automatic chatbot knowledge acquisition method from online forums is presented in this paper. It includes a classification model based on rough set, and the theory of ensemble learning is combined to make a decision. Given a forum, multiple rough set classifiers are constructed and trained first. Then all replies are classified with these classifiers. The final recognition results are drawn by voting to the output of these classifiers. Finally, the related replies are selected as chatbot knowledge. Relevant experiments on a child-care forum prove that the method based on rough set has high recognition efficiency to related replies and the combination of ensemble learning improves the results.