Rasa in AWS ECS
Choose EC2 > 16GB (preferable is 32GB)
Rasa needs TensorFlow, spacy models that have a considerable footprint (~8GB) so I would suggest to use at least 16 GB memory.
The first week of this year has been spent creating a page for The Intent Engine on TrailSix.com. I started by writing the vision and the planned roadmap. The roadmap includes the NLP, NLU, Dialogue Management and Action components that make up the engine. The next section includes icons, which allude to each I/O interface I plan to develop and integrate into the engine. For example, CLI, web, slack, fb messenger, apps, etc. Maybe one day Neuralink will be one of the I/O interfaces!
Blog and Vlog posts will consist of my progress as I build The Intent Engine and the interface integration. At this point I’m torn as to where to capture the ideas. I like writing here on my personal blog because I feel it captures my thought process. For example, I write about philosophy books that I read and books about how our minds work. All of those concepts play a role in how The Intent Engine is developed. Privacy and security are sure to play large parts in the development. I believe that your data is your data. It should be fully controlled by you the owner. You should have first say as to what is stored, what is not and if it’s stored for how long. All of these are important topics to think about when designing The Intent Engine and I plan to cover them all and my thought process behind them here. For now, I think this is the right medium and place (blog.joebarger.com) to capture my thought processes. I created The Intent Engine project at my other website TrailSix.com because I view that as more of a business. I view TrailSix as a parent or holding company for all of my tech ventures, which are subsidiaries or business units. The Intent Engine just happens to be what I’m working on at the moment. I’ve thought about giving The Intent Engine its own domain since the project name is quite long, but it doesn’t seem like I need it yet. I believe that change is constant so when the time comes I might do that. Enough about that. Let’s get back to the topic at hand.
The Intent Engine project is more than just Natural Language Processing though. It’s about understanding people, their thought processes, the way they identify with things and others, our psychology, how we learn, etc. There are a lot of abstract concepts combined in this project and also in the blog posts as well. The “end” product will be the intent engine, but I don’t know that development will ever be completed as we’re always learning more about ourselves, each other and the world we live in. So it’s a really fun endeavor with a very large scope that lets me combine and explore my major interests.
Speaking of how we think, I thought it important to focus on a few intents to get started. NLP supervised learning still requires a lot of manual input for training an ML model so it’s important to focus on specific tasks until the code is capable of expanding. Therefore, I thought it was best to focus on the 6 steps in the buying process. Each of those is a different intent that occurs when you are in a different mindset and potentially requires different sets of actions. When it comes to coding we’ll actually start with the search intent out of all six.
|Buying process step||Intent||Action|
|Evaluation of alternatives||compare|