
AVIOS
Conversational Interaction
Conference
CONNECTING WITH COMPUTERS - SPEECH, TEXT OR BOTH?

In the Heart of Silicon Valley
SAVE THE DATE - Feb 5-6, 2018
NEWS
From the July 2017 issue of LUI News (on the Language User Interface):
Facebook Messenger Platform 2.1 supports enriched automated conversations
Facebook’s Wit.ai provides built in natural language processing API for the release
Facebook has launched Messenger Platform 2.1, which includes new and improved features that allow businesses to develop new ways to connect with their customers and enrich their conversational experiences on Messenger. Facebook’s Wit.ai separately announced the launch of built-in natural language processing (NLP) for Messenger, launching with the Messenger Platform 2.1.
New release
When 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 first version can detect the following entities: hello, bye, thanks, date & time, location, amount of money, phone number, email, and a URL. Facebook said this is a first step in bringing NLP capabilities to all developers. The Wit.ai discussion at the end of this article contains further information.
A new handover protocol enables businesses to create multiple experiences within a single bot. For example, a business can use this to move seamlessly from an automated (bot) to a one-to-one (human) conversation. Handover protocol makes it possible for multiple service providers to power a single Messenger experience. As part of Platform 2.1, handover protocol is now available globally as an open beta. E-commerce merchant Sephora will be implementing handover protocol soon, incorporating a combined experience from chatbot service Assist and social technology company Sprinklr.
The new release makes payments easier to set up and more seamless for people to check out with a new SDK for the webview payment process. Customers can pay using a one step process through Messenger webview.
Facebook has launched five new Call to Action (CTA) buttons that businesses and developers can add to their Page to drive people to their Messenger experience. In addition to the currently available “Send Message” CTA, businesses and developers can select from the following options: Shop Now, Get Support, Get Updates, Play Now, and Get Started. These descriptive CTA options let people know what to expect when they start a conversation with a business.
Several other features make using Facebook by businesses easier.
Wit.ai
Wit.ai said that testing and learning from the their Bot Engine beta, the company determined that it made the most sense to refocus on pure NLP to make it accurate, reliable, and scalable for everybody. The company is thus sunsetting Bot Engine and will stop serving requests on February 1, 2018.
Bot Engine was trained with a small number of “Stories.” Stories are example of conversations that contain a user’s and bot’s messages and the resulting bot actions. The approach allowed launching a bot based on machine learning, but which overfit the data and essentially acted like a rules-based approach. Adding additional cases over time made the approach more general, evolving toward a more flexible machine learning solution.
Wit.ai described the main problem with the Bot Engine, however, as its focus on text-only interaction. Since Facebook Messenger is always used with a visual interface and, in many use cases, a dialog does not provide the user experience you can get with a Graphical User interface (GUI)—constant visual feedback, ability to modify previous choices, etc.
Since the launch of Bot Engine in 2016, the company has seen its community grow from 20,000 to more than 100,000 developers. Most of them build bots for Messenger, Slack, Telegram, and other platforms and use the company’s NLP API rather than the Bot Engine. Currently, more than 90% of the Wit API calls are coming from the NLP API.
In the last year, Wit.ai said, the ecosystem has shifted towards a mix of NLP and GUI elements to produce a user experience on par with native and web apps. For instance, Messenger has introduced quick replies, menus, and web view. As a result, Bot Engine and its emphasis on text-only bots has become somewhat obsolete.
Wit.ai said that it believes almost every bot should use NLP in some way to create “delightful experiences.” Currently, if you are leveraging an NLP API (/message) in Facebook Messenger, it is an additional layer that adds both latency and complexity. Wit.ai is making this easier by integrating Wit directly into the Send/Receive Messenger API.
As noted, when Built-in NLP is enabled, it automatically detects intent and content that is passed to a chatbot. The first version is limited to the functions mentioned in the introduction to this article.
The company said it is working on making NLP work well at scale, be it conversational or hybrid NLP/GUI. The company has scaled its /message endpoint to make it easy to programmatically interact with Wit.
Moving forward, the company is working on:
-
Improving the quality of their NLP by leveraging cutting-edge algorithms developed at Facebook;
-
Making it easy to share, reuse, and collaborate on entities from the community; and
-
Helping other platforms leverage the company’s NLP API behind the scenes.