Chatbot Architecture for a Footwear E-Commerce Scenario.

The amount of information accessible to businesses grows as more individuals get access to the internet. This information can be used by businesses to make corporate decisions that can highly affect them. E-commerce is the result of businesses starting to export their operations to the internet as a result of taking advantage of this reality. To help a footwear e-commerce platform keep up with the current demands of the market in terms of accessibility and customer service, a chatbot was created to help on this request. To build this chatbot, datasets were created leading to Intent Classification and Named-Entity Recognition pre-trained models, utilizing Bert-Large, getting fine-tuned with those datasets. These models achieved a F1-score of 0.90 and 0.88 respectively in their tasks. The Sentiment Analysis functionality of TextBlob was also utilized to help the chatbot comprehend user’s text polarity to reply appropriately.