Can I build my custom AI assistant using dialog flow?

Question

I am so confused. I want to build a chatbot like Siri but for my own tasks. It should be able to : - search on the internet and get answers of questions . - give people specific information daily. - discuss with people some scientific phenomenas.

I can't determine what platform should I use to build this chatbot . I thought about using Dialog flow but I can't figure out if it will give me the ability to do that or no. Also I thought about using tensor flow but I think that it will take a very long time so I was wondering if I can achieve what I want with using some thing like dialog flow and not to build it from scratch?

Answer 1

Actually in my opinion, DialogFlow is the best option to build an assistant, it is really easy to build a chatbot to save reminders, to check the weather or to have a simple conversation. DialogFlow has a really powerful tool called webhook that uses Cloud Functions to do the real programming, for example, call google APIs, such as Translate API, or insert data in your Cloud SQL database.

Also, DialogFlow use Machine Learning algorithms to understand the customer, for example, if the client says: "What's the weathe in Barcelona?" It will answer the question correctly.

Another great feature is that it is integrated with multiple technologies, such as, Google Assistant, Amazon Alexa, Cortana, Telegram, Line, Facebook Messenger, etc.

I recommend you to follow this tutorial.

Answer 2

Luis from Microsoft and DialogFlow from Google allow you to build models for natural language processing. These models need to be trained. So the answer is "no", out of the box, these tools do not "search the internet" to discover answers for your intents.

What you need to do is figure out how to train a natural language model, and integrate search data therein. This is bleeding edge AI. And this really is your question; "How do I integrate search with NLP and/or chatbot?"

Both Google and Microsoft let you hook into search. You do not need the dialog tools to do this; you can just pass the query text to do the searching (and let the engine use both ML and heuristic methods to rank results). You mentioned IBM Watson and this is a tool that uses ML modelling to try and answer QnA questions. The Google competition is DeepMind. You can check out those yourself.

But I believe curated content is often the way to go. Tools like Microsoft's QnA Maker let you build these types of applications very easily with little programming required. You can also look into the Azure or Bing search APIs.

And if you are looking to start with a bot from template, there are tons of examples on GitHub for Azure Bot Service and Actions-on-Google. Some even integrate with search and QnA tools. :-)

(And here is the disclaimer. I work for Microsoft. My views do not represent that of my employer.)