Bots now build themselves! We’ve created a bot generator allowing the automatic creation of Q&A chatbots based solely on the upload of a policy document. Having a fully functional bot now takes 9 minutes instead of 9 weeks. You'll soon be able to test it yourself!
Q&A Bot Generator (closed beta)
We know training bots can be hard. From now on, we tell you everything you need to do to build the best trained bot! Our brand-new analytics panel allows you to understand the performance of your training datasets by automating your benchmarking process. We also give you live tips on how to improve performance!
On the Monitor tab, the Training Analytics section helps you to build a great dataset for your bot. These analytics are only available for bots with at least 5 intents and at least 30 expressions per intent.
Your dataset (that is, all the intents and entities that you created and trained) is a fundamental element of your bot. If your bot isn’t well-trained (meaning your dataset isn’t well-structured or is incomplete), your bot won’t be able to correctly understand messages from its users, resulting in a disappointing conversational experience.
Learn more in the documentation and discover these features:
✨ Live benchmarking
✨ Automatic calculation of 4 key metrics
✨ Interactive confusion matrix
✨ Live tips on how to improve bot training
You don't need to plan extra time for pushes to production or upgrades anymore. We're glad to present our new powerful versioning system! It allows an easy maintenance and updating of bots, as well as a powerful development environment manager. Putting into production is now instant and secure for your entire team.
To enable versioning for your bot:
Go to the Settings page for your bot. Under VERSION SETTINGS, activate versioning. At the top of the page, you’ll now see a dropdown to create and switch to different versions.
A version is a package of your bot training dataset and skills. Each version is independent of the others and can be managed individually. For example, you might want to create a new version prior to major updates to your training dataset or skills. Additionally, separate versions may be useful if you want to leverage two variations of the same core bot.
The persistent menu
Bot Connector now allows you to configure a persistent menu for Facebook Messenger and the Webchat directly on the platform.
The persistent menu allows you to have an always-on user interface element inside your conversations. This is an easy way to help people discover and access the core functionality of your chatbot at any point in the conversation.
For Facebook Messenger, you need first to activate the Get Started button.
Get Started button for Facebook Messenger
Bot Connector now allows you to configure the "Get started" Facebook Messenger button. The welcome screen is the first thing people see when they encounter your Messenger bot, and includes information that allows a person to learn about your bot and what it offers.
The welcome screen includes a get started button, which starts a conversation with your Messenger bot when it is tapped. When a person taps the get started button, the message that you've configured in Recast.AI will be posted into the conversation, and your bot is then granted permission to send messages.
SAP CEC History
The conversation history that was transfered to human support agents through SAP CEC is now complete and accurate.
We've added the following new features to Webchat:
You can restrict messages from users to 512 characters or less. For example, you may want to do this if users tend to add a lot of details that obscure the intent of the request.
Go to your Webchat configuration in the CONNECT > Your webchat > 5. Global Settings
In the 3. Message settings you can configure the way the text appears in the text box.
Reordering Actions in the Bot Builder
On the Actions tab in the Bot Builder, you can now easily reorder message groups by drag and drop.
If you've added the languages English, French, German, or Spanish to an intent, and enter a new expression for the intent in any of those languages, the platform now automatically suggests additional expressions in those languages. You can quickly and easily add the suggested expressions to the intent. This increases the size of the training dataset, enabling your bot to understand its users more precisely.
For improved monitoring of your bot’s usage, as well as to simplify training, you can now filter your logs by intent, language, date, and strictness. You can also download all your logs as a .csv file.
No published changelogs yet.
Surely RecastAI will start publishing changelogs very soon.
Check out our other public changelogs: Buffer, Mention, Respond by Buffer, JSFiddle, Olark, Droplr, Piwik Pro, Prott, Ustream, ViralSweep, StartupThreads, Userlike, Unixstickers, Survicate, Envoy, Gmelius, CodeTree