Microsoft has made available Bot Framework Preview to developers and Google has purchased API.ai, a bot engine with many integrations.
Earlier in the spring Facebook opened up the Messenger Platform and a bot engine hosted by Wit.ai. Two months later, they released a new version of the engine and more than doubled their developers from 20K to 45K. It is not sure how many of these are actually building chatbots integrated into Messenger and how many are just testing the waters. One thing happened though: the large influx of developers polluted the training dataset of built-in entities such as wit/location, making them inaccurate and requiring more work to clean it up.
Since April, Microsoft has made available their Bot Framework, currently in preview mode. Developers can create chatbots for the Bot Framework in C# and Node.js, applications that are deployed and run on Azure. Unlike Facebook’s bot engine which can be used only inside Messenger, bots created with Bot Framework can be integrated both with some Microsoft applications - Skype, Email, Office 365, GroupMe, SMS - but also a number of other applications not owned by Microsoft, such as Facebook Messenger, Kik, and Slack. They can also be embedded in web pages.
The chatbots created with the Bot Framework can use Microsoft Cognitive Services, a set of APIs that can provide vision recognition, emotion analysis, facial recognition, video detection, language understanding, search and others. These chatbots rely on Microsoft Language Understanding Intelligent Service (LUIS) to interpret natural languages. Supported languages are: English, French, Italian, Spanish, and Chinese. The Bot Directory currently contains 55 chatbots created by various companies.
After acquiring Apigee, an API management company, Google has bought API.ai, another bot engine. API.ai is similar to Wit.ai, uses the same concepts – entities, intents, actions – but it is integrated with many applications, including Amazon Alexa, Cortana, Kik, Facebook Messenger, Skype, Slack, Twitter and others. They have SDKs for many operating systems, platforms and languages, such as Android, iOS, Xamarin, Unity, Cordova, C#, C++, Java, JavaScript, PHP, Python, Ruby. They also mention they have processed over 3B requests so far.
It is not clear why Google bought API.ai, considering they have better natural language understanding services. Perhaps this acquisition has to do with the large number of integrations with other applications and the 60,000 developers using it. This would give Google a userbase where they can bring their chatbot technologies such as Assistant. A preview edition of the Assistant was made recently available with Allo, a new chat application from Google. Assistant understands natural language and attempts to solve everyday tasks such as providing news, answering questions, translating phrases, booking flights or restaurants, work with emails or contacts and others.