There are 3 technical achievements that our team is would like to highlight.
Firstly, major chatbot frameworks are only introduced in the past year or so, and have been rapidly evolving with significant updates published during the time of our development. We are able to understand the chatbot frameworks that are available in the market, and objectively weigh the pros and cons in deciding which framework we would work with and produce our chatbot service. Knowing how the Watson conversation platform functions, and taking into account its limitations, was key to the success of developing our solution.
Secondly, our team decided to use Node.js as the back-end programming language for the chatbot service. Despite having no experience using this language, we thought that it will be a good learning experience and will be useful for our future development purposes. We are glad to be able to learn the basics of Node.js and hopefully apply this new knowledge to future projects.
Lastly, we automate a lot of the data collection required to import into the Watson Conversation workspace that trains the chatbot on recognising intents and entities. We wrote scripts that convert the JSON response from NET-A-PORTER APIs into CSV format files that we can import into the workspace. Also, misspellings and name variations of entities were produced via scripts and include scrapping results from an online typo generator using Java Jaunt.