Daniel_ (5) [Avatar] Offline
Hi guys,

I would like to know if you are going to do a serious chapter about chatbots. Currently, I am working creating a chatbot for an important company, and I have found that the best framework combination to work on it is to use nodeJS, microsoft bot framework, and api.ai (Google conversational engine instead of LUIS microsoft engine), and stanfordNLP server to get some extra info like pos-tagging and lemmas.

Are you going to include anything about this new NLP tools that enterprises are using to create their chatbots?
428125 (18) [Avatar] Offline
Absolutely! We have a chapter dedicated to the scalability of the entire NLP pipeline using tools like TensorFlow, Keras, gensim, and SpaCy. And another chapter that describes the chatbot built around those tools. But we will not discuss distributed computing platform or AI-as-a-service NLP-as-a-service options, because those services are often opaque black boxes that don't help build understanding of the underlying NLP algorithms. We want you to be able to build your own service and customize the NLP pipeline for your needs. It helps to see the source code when you are doing that.

The tools and frameworks we use for our chatbot are scalable and stable for production use at corporations where chatbots are core competencies. As you can imagine, your stack is not the only effective architecture. And our book is centered on the state-of-the-art tools that have python APIs, since these are the most productive tools to use in an environment where code maintainability (readability makes it easier to work on someone elses code) and reliability (readability improves test coverage and code correctness) is the top priority. Many leading companies like Google place high value on the productivity of their developers and devops staff and use a python-centered stack for applications like NLP pipelines on many of their teams, especially those that use Tensor Flow and SpaCy, like we do.
53737 (2) [Avatar] Offline
Please, what the benefit of learning NLP theory and algorithms as oposed to using ready tools like Micosoft bot framework and api.ai if the objective were to build a comercial bot? Is it not the case that MS and Google invest ahuge amont of money in their products making them the best possible ? TVMUCH, RConte.
468265 (3) [Avatar] Offline
Hello, I work for a professional chatbot company, similar to api.ai, luis, etc... and I am finding this book extremely interesting, it is even helping us improve already.

Even thou I understand how people want to learn already working NLP services, I think they can solve their needs by reading those services documentation, watching youtube videos and learning coding skills like clean coders.

For curiosity, what would you expect to find in that non-existing chapter about existing chatbot/NLP services?

In my opinion it wouldn't hurt to have a chapter about existing services, I would happily read it, but that is a short term knowledge that will be useless in a very short time, while understanding the underneath NLP from its classic form to the modern form, as the book is doing, is a knowledge that will be useful for me for many many years.

Thanks for such a great book already, please feel encouraged.
447276 (1) [Avatar] Offline
I agree with the authors. Using APIs of existing frameworks can be learned by reading the documentations or YT videos or plenty of resources out there. Having a deeper intuitive learning of how it is built from scratch, however, is a totally different game and should be focussed on.