NAP

The front-end chatbot interface is built purely with HTML, CSS, JavaScript, AJAX. A bit of bootstrap is used as a UI framework. User queries are passed on to the REST API and responses (JSON format) are displayed, which includes product recommendations and customer care information.

The REST API encompasses the our application that uses the IBM Watson services shown in the architecture diagram. This application is exposed via an end-point hosted on Amazon Web Services (AWS) using API Gateway, wrapped in AWS Lambda.

Watson Language Translator is used to detect languages in the user query. Currently, only English language queries are supported and interpreted via the chatbot service. German, French, and Mandarin language queries will result in response to change the website language as it supported by NET-A-PORTER.

Watson Conversation is used to detect intents and entities in the user query. Product recommendation intent will pass on the entities into our back-end parser to generate results using NET-A-PORTER Listings and Designers (LAD) APIs. FAQ intent will pass on the user query into Watson Retrieve and Rank (R&R), which will fetch the response from the knowledge base in the information cluster within the (R&R) instance. Watson Document Conversation is the service used to covert NET-A-PORTER FAQ information to import into the information cluster.