Interactive Applied Graph Chatbot with Semantic Recognition

Companies and small medium businesses (UMKM) need to interact with customers to increase their customer retention rate. One of the ways is to use chatbot. Aside from being cost-effective, this method is also very effective and very easy for companies to use. To make an easy and effective chatbot requires a combination of two scientific fields, artificial intelligence and software engineering. This study has the following features. 1) Affective sentiment analysis, this feature is inspired by Russel's Circumplex Model. Adjective words will be mapped in a matrix with values based on the Russell Circumplex Model. This model will pay attention on the adjective words, polarity, and affection degree of a sentence. 2) Conjunction sentiment analysis, consider of free way of interaction nowdays, a sentiment analysis system need to determine the sentiment value in multilevel sentences. This multilevel sentence has one or more conjunctions. This conjunction sentiment system will break sentences based on conjunction. The fractional sentence will be processed by affective sentiment analysis. The system then considers the nature of the conjunction to determine the sentiment of the whole sentence. 3) Graph Chatbot, this feature is used to make it easy for companies to modify and generate their bots. This bot will interact with customers like humans, based on a graph chat map that has been created. Chatbot graphs were developed using javascript library Vue js to make it easier to manipulate the visual graph. The system produces satisfying accuracy. Graph chatbot can handle procedural conversations very well. The flow of conversation in accordance with the graph that has been defined. Graph chatbot has 100% accuracy and successfully responds to all user conversations. The sentiment method has 63% accuracy.