Neural networks based NLU commoditized through Google SyntaxNet
What does WebRTC have to do with SyntaxNet?
Conceptually it is the same story all over again. Google is again taking a technology that was until now in the hands of the rich and experts making it a feature in any application.
On the product side WebRTC applications can benefit a great deal from this, more on that further in this post.
Starting from the beginning. What is SyntaxNet?
SyntaxNet is a Google technology (surprise surprise) based on neural network technology (TensorFlow, that’s Google again) that is the foundation for Natural Language Understanding (NLU) and it has just become open source.
Why is it a foundation and not a complete solution?
Because you need to train it for a specific language. Google is closing this gap by including in this open source 2 more things:
- The English parser named Parsey McParseface making it a complete solution for English
- The tools to train SyntaxNet for other languages
Similar to WebRTC, SyntaxNet is a technology, not a complete solution or service.
Read this post to understand the complexity and the method used to solve the problem.
Why open sourcing SyntaxNet is similar to open sourcing WebRTC
Back in 2011 Google came out with a humble, low profile announcement saying it is making WebRTC available as open source.
This announcement is changing communications as we once knew it:
- Making it possible for developers with no VoIP knowhow and assets to add communications to their application as a feature
- Moving the cheese for service providers and communication technology vendors as now there is less need for their equipment and services
- WebRTC is used in ways people didn’t think of before
- An ecosystem has been built around WebRTC
- VoIP has been democratized and commoditized
Fast forward to May 2016. SyntaxNet will do the same to the complex technology of NLU. Until today, building Siri, Cortana, Google Now and Amazon Echo was a complex technological challenge possible to be achieved only by large companies. With SyntaxNet becoming a free to use open source, Google is doing to NLU what it did in 2011 to communications:
- Making it possible for developers with no NLU knowhow and assets to add NLU related features to their applications
- Making it possible to build NLU related applications and services we never thought of before
- Creating an opportunity for an ecosystem to be built around NLU
- Allowing small companies to move the cheese currently in the hands of the incumbents
VoIP has been democratized and commoditized, now Google is doing the same to NLU opening boundless opportunities for integrating this capability into existing applications such as home automation, IoT, customer service, banking, healthcare, education… and communications.
WebRTC and NLU
Both WebRTC and NLU are technologies. They can be combined together. A simple example, already somewhat possible on mobile phones is to control the WebRTC client (Call X, add Y to conference…).
Another implementation can be related to meetings and customer service interactions archiving and manipulation of the recorded session for extracting highlights, summarizing action items or customer inputs and finding the session recording or searching within the recording itself.
These are just 2 simple examples. The availability of an open source NLU framework further expands the opportunities of applications that use these 2 technologies to areas we can’t imagine but developers will create this long tail of applications as they have done with WebRTC.