Customer Support Chatbot for Electronic Components

Customer satisfaction is a key metric for any company providing products or services. A great deal of customer satisfaction is directly related to the support provided by the company to its customers. This paper describes an automated solution to achieving good customer support. The focus is on an e-commerce industry selling various electronics components such as sensors, micro-controllers and actuators. One of the key challenges is in the large variation in the types of queries and possible solutions. Different NLP based data exploration techniques are employed before moving on to exploring different model architectures. The solution described in this paper is based on a Seq2Seq LSTM model with Attention. The data from a private company is used and an evaluation accuracy of about 0.90 is achieved. However, this could be misleading since the dataset is very small, and additional evaluation metrics need to be measured and compared.