Use Rasa for Your Next Hackthon Project

I looked at Rasa project in early 2019 for an internal Hackthon project. I guess, by that time, the conversion AI was not that popular as earlier years, and since then, there was not that much blogs and news about this platform. I don’t have a data to prove that, just by my impression while search online. So use your discretion for that statement.

In July of 2020, internally we have another hackthon project. One of our team member (the boss) proposed a conversion AI bot to answer questions about information between a guest and a Vrbo host. Those questions are pretty common such as

1. the wifi password
2. Keypad passcode
3. Cancellation policy
4. earlier arrival policy
5. check-in / check-out instruction
6. how to use washer /drier, nest thermostat, and so on
7. … …

I have used Airbnb in the past year, I can be pretty much related to those questions. Normally, I will stand in front of the door for a few minutes to find out the keypad passcode.

For our hackthon, my intention are that:

1. learn something new
2. don’t need to spend too much time on it
3. don’t need to stress out
4. still be able to have a concrete and useful demo, by the end of day.

With those being said, I proposed Rasa as the tool to build our chatbot. It turned out perfectly achieve our goal and my intention.

I will not walk you through how we build our bot. It is a prototype and built by engineers who don’t have any pre-knowledge about natural language processing, Rasa and whatsoever.

Here are a few things I want to point out while you take a look at this Rasa for your project.

the best online resource

Most of them were quite out-dated. I think this is the most updated and accurate tutorial about Rasa by Rasa team.

Rasa Masterclass: Developing Contextual AI assistants with Rasa tools on youtube

You can also find its ebook here.

Also the source code along with this video tutorial.

You might notice that the source code was last updated 6–8 months ago and it didn’t represent the latest Rasa, unfortunately. and you will notice the lecture8 (and final) of the source code didn’t match the video content.

In the meantime, if you search the medicare locator project with Rasa’s github repo, you will find this project:

RasaHQ/medicare_locator

This project works but only for older Rasa version. and It didn’t work for Rasa X.

I kind think Rasa X is the best that Rasa open source is offering. Buggy but work well for your hackthon project purpose.

Anyway, if you are trying to get this project up and running with the latest Rasa and Rasa X with useful instruction, you can check my fork out:

liangjun-jiang/future-travel-hackthon

I could have sent a PR, but I do realize no one is really paying attention on those Rasa’s project’s PRs.

Rasa just received $26M funding on June 26, 2020. I assume it will get even better in future.