Integrating machine learning and open data into social Chatbot for filtering information rumor

Hsu, I-Ching; Chang, Chun-Cheng

Social networks have become a major platform for people to disseminate information, which can include negative rumors. In recent years, rumors on social networks has caused grave problems and considerable damages. We attempted to create a method to verify information from numerous social media messages. We propose a general architecture that integrates machine learning and open data with a Chatbot and is based cloud computing (MLODCCC), which can assist users in evaluating information authenticity on social platforms. The proposed MLODCCC architecture consists of six integrated modules: cloud computing, machine learning, data preparation, open data, chatbot, and intelligent social application modules. Food safety has garnered worldwide attention. Consequently, we used the proposed MLODCCC architecture to develop a Food Safety Information Platform (FSIP) that provides a friendly hyperlink and chatbot interface on Facebook to identify credible food safety information. The performance and accuracy of three binary classification algorithms, namely the decision tree, logistic regression, and support vector machine algorithms, operating in different cloud computing environments were compared. The binary classification accuracy was 0.769, which indicates that the proposed approach accurately classifies using the developed FSIP.