Neeraj Chaudhary, Solutions Engineer, Symbl.ai
Symb.ai is an API platform to analyze voice, text, and video conversations in real-time using artificial intelligence and machine learning. You don’t need to build any machine learning models. Symbl.ai can contextually understand conversations for your applications. The platform will also work on pre-recorded content (async), not just real-time conversations.
They’re also a sponsor of TADHack Global 2021 in September, more coming on that very soon.
The contextual understanding of a conversation is a significant acceleration for developers as training a model requires gathering data, cleaning it so the data trains the model accurately. With Symbl.ai you can skip this step, which for a hackathon is really handy 🙂
Neeraj walks through how you can quickly can get up and running with a great demo using their ‘symbl-twilio-video-react’ repo on their github page. A video application that uses Symbl APIs, built with Twilio video SDK and React.js
I particularly appreciate the explanation on initial set-up as often those first couple of hurdles, though straight-forward once you’ve done it once, can catch you out. Yes, I know, read the documentation, but who does 😉
Neeraj then fires up the app, explains port usage, the javascript code, and connectors. It’s a nice walk-though to get you hacking fast. Thank you Neeraj.
Thanks Neeraj for a great walk-though, you make it look so easy 🙂 I have a few questions:
1) What are some of the AI insights available though your API?
2) For getting started your step by step guide is great. What are the most common mistakes people make in getting started.
3) What are some of the new Symbl.ai features you’re using at the moment?
4) I noticed a repo for Symbl.ai on top of Firebase, which is used quite often at TADHack. What’s happening there?
1) What are some of the AI insights available through your API?
Symbl’s can provide you insights corresponding to any conversation data, let it be audio, video, or text , recorded files or real-time stream . Below are some of the insights that you can get through Symbl corresponding to your conversation data :
Actions Items
Follow-ups
Questions
Sentiments
Topics discussed during the conversation
Entities
Analytics: Like WPM, speaker ratio, etc.
There are several advanced features like automatic speaker separation etc are available as well ,more information about these and other insights can be found in our documentation( https://docs.symbl.ai/docs/ ).
2) For getting started your step-by-step guide is great. What are the most common mistakes people make in getting started?
Our API documentation is very simple and straightforward and anyone can get started within 5 mins .
I would generally classify it as a 3 step process :
1. Create a free account on Symbl homepage and note your APPId and APPSecret (don’t worry you will get plenty of free credits ).
2. Submit an Audio/Video/Text file using our API (POST and PUT call) and note the Conversation ID received in the response of the API call.
3. Retrieve the AI insight and more data using the conversation ID.
Even though the API’s are really simple, You will need to make sure that you are passing all the required parameters while making the REST API call . A good way to avoid any mistakes there is to download our postman collection( https://docs.symbl.ai/docs/developer-tools/postman/ ) and make API calls through that , that way you can avoid any unforced errors.
To get started, I would recommend you should start with our Async API’s ( https://docs.symbl.ai/docs/async-api/introduction ) and can see how powerful our API’s are.
For real-time streaming, the guides are good for basic setup and the only things you need to consider are sample rate , encoding of the stream, and other configuration parameters that are there in our docs.
But if someone is planning to build a production-level integration I would recommend considering factors like speaker separation, no of speakers, number of channels, handling audio stream, protocols etc . Obviously, on our end, we can handle any permutation and combination of streams, but from an integration standpoint, I would recommend having a plan on how are you going to handle your streams and then proceed. Please feel free to reach out to support@symbl.ai for any queries and assistance.
3) What are some of the new Symbl.ai features you’re using at the moment?
Symbl is always coming up with cool new features and one of our most liked and talked about features are :
Trackers:Trackers allow you to track the occurrence of certain keywords or phrases in a conversation so you can identify emerging trends and gauge the nature of interactions. You can define keywords or phrases in a Tracker and Symbl will return messages that contain the same or contextually similar phrases . More information can be found here (https://docs.symbl.ai/docs/concepts/trackers)
Summarization: It can summarize a whole conversation into 2-3 sentences.
There are other new features like these that are getting released all the time. Please keep visiting our docs for other latest features.