Wit.ai launches built- in NLP for Messenger and shuts down Bot Engine

July 29, 2017, 5:33 a.m. By: Prakarsh Saxena


Wit.ai – the start- up acquired by Facebook a couple of years ago, has finally shared the news of the launch of a built- in NLP for Facebook Messenger, launching with the Messenger Platform 2.1. It focuses on pure NLP to make the interaction with the bot using it to be more reliable, accurate and scalable for every developer. This also implies that they need to take back the Bot Engine as Facebook will be rolling the NLP into their platforms. They will also stop serving requests to the associated /converse endpoint on February 1, 2018.

What’s with the Built- in NLP?

Quite a lot of developers around the world have tried to develop their own bots using any NLP API available. Currently, any leveraged NLP API adds an additional layer of latency and complexity. This had given rise to many developers questioning how to integrate Wit.ai NLP API into their bots for messengers. Answering them, they have finally launched the API directly integrated into the Send/Receive Messenger API.

When the Built- In NLP is enabled, it automatically detects meaning and information in the text of messages that a user sends, before it gets passed to the bot. The Beta version launched can detect standard entities like ‘hello’,’bye’,’thanks’, date and time, location, amount of money, phone number, email and a URL. Wit.ai says this is the first step in bringing NLP capabilities to every developer.

Shutting down the Bot Engine

In order to aid the developers to build text- based conversational bots, Bot Engine was created by Wit.ai back in 2016. It was a push in the direction to promote the bot as well as the NLP tools ecosystem. Back then, messaging platforms did not have GUI elements. Bot Engine was hence thought of to replace forms with proper conversations, albeit without a GUI.

The ecosystem then shifted towards a mix of NLP and GUI elements to produce a better user experience with native and web apps. Messengers, for e.g., introduced quick replies, menus and even web view. This reduced the need and emphasis on text-only bots.

Upon reviewing the top apps using Stories, most apps used 1-turn stories. These FAQ-like apps were easily achieved with their NLP endpoint (/message) because it’s about understanding the question and then mapping the user intent to an answer that can be stored outside of Wit. These apps didn’t need the power and complexity of Stories (having to do 4 calls on average to /converse to get to an answer as opposed to 1 single call for /message). The efficient migration was to use /message and code on the developer’s side to manage the conversation and link to the answers that would be stored in their database.

Since the Bot Engine was launched, the community staggeringly grew from 20k to more than 100k developers in a very short time. Most of the developers built bots for Messenger, Slack, Telegram and other platforms. But then, more than 90% of the Wit API calls were coming from the NLP API. This massive diversion of call traffic has led them to finally decide on the depreciation of the Bot Engine and the Stories UI, and will stop serving to requests associated with the /converse endpoint on 1st February 2018 to allow developers to migrate the affected apps to the better API. You can check their GitHub tutorials here.

Wit.ai have been working to make NLP better for everyone. They are still aiming to improve their NLP quality by using highly efficient and innovative algorithms developed at Facebook, and pushing others to leverage their NLP API for a smoother experience to their users.