Last year, I did the integration of WhatsApp business account for my Organization, which enabled us as one more source to serve our users. During the requirement gathering phase, I realized to make the bot smart we need NLU, which will take a lot of time, and had 0 experience and knowledge of NLU and NLP. I was not even sure whether this will add value to the business but liked it this was interesting. After a few discussions, the decision has been made let’s take a simple bot that works on string manipulation and with no NLP, and let’s see how it will perform and take a decision.
The bot was taken live to serve only one query, anything other than that query the bot will reply thank you. The business requirement was fulfilled but as a developer, I know there was a scope to improve.
We have a Hackathon yearly in our organization. I was sure with the topic I’m going to pick this Hackathon that is around this feature. After discussing with my Hackathon team. We concluded there were 2 areas to improve in this feature.
- NLU
- Insightful data extraction from message logs
After hustling for 30 hrs, we successfully integrated Amazon Lex into our bot, trained it for that single query, and able to replace our current string manipulation. One of our team members was from the analytics team. He did an awesome job of creating the panel from the message logs.
