Envec83

I am a programmer/entrepreneur from Brazil, and I will be spending some time in San Francisco. I am looking to expand my network here and make some friends, and I figured a good way to do that is to work for a startup.

I am willing to work for free as long as I have fun doing it, and I have a lot of experience with:

- SEO/Content Marketing - Email marketing - AdWords, Facebook Ads, Display ads - Programming (pretty much anything from Assembly to JavaScript)

You can reach me here: danielscocco@gmail.com Personal blog: http://danielscocco.com

Let's say we want to develop a chatbot whose objective is to be able to talk to people about cars (i.e., "What do you think about BMWs?", "What is the fastest sedan in the market?", etc.). I am narrowing the domain on purpose.

In order to build such a bot, some approaches come to mind:

1. Hardcoded answers: use NLP to understand the meaning of the user input, go through defined rules (if-else basically) to choose the appropriate answer from a set of hardcoded, pre-defined ones.

2. Machine learning on existing conversations about cars: if you have access to lots of conversations about cars, you could use machine learning to train the bot to understand the meaning of the user input and to formulate an answer based on the data.

3. Machine learning on generic information about cars: same idea as above, but using generic infomation about cars (i.e., wikipedia, magazines) to formulate answers. NLP would need to be used both to understand the user input and to formular syntatically and semantically correct answers.

Questions: Are these approaches indeed viable? Are there others? Are there examples of bots using 2 or 3 out there?