The feature enhancements in this release empower bot developers to design intelligent bots using options like context switching and improved machine learning training models. Enterprises have better control over their bot deployment strategy by taking advantage of smart bots and applying redaction techniques for information security compliance.
Enhanced context switching and interruption handling
In this release, the bot’s context switching abilities were enhanced to handle interruptions in conversations efficiently. Developers are provided options to control such experiences. Apart from improving ‘Hold & Resume’ options at a bot level, we now allow developers to customize the “Hold & Resume” behavior for specific tasks and nodes. You can also add conditional exceptions between tasks with the ability to pass contextual data between them. Using ‘Follow-up Intents’, the bot will now capture all unattended interruptions from a conversation flow and make them accessible to the bot as an array.
Help link: Interruption Handling and Context Switching
Introduced Smart Bots: Reusing and extending bots
A new bot type, Smart Bot, has been added to the Kore.ai Bots Platform in this release. Smart Bots enable enterprises to expedite the bot building process by allowing new bots to inherit the configuration and tasks of previously built bots. The child bots can then extend the inherited functionality by adding further tasks and knowledge queries, as well as localize relevant parameters using variables. All updates made to the inherited bot elements in the parent bot are automatically passed on to all connected child bots.
Implemented redaction option to secure sensitive user information
Kore.ai now allows you to redact any sensitive information that users share with bots. When you enable redaction for an information type, you can transform its value into a pattern that contains garbled content so critical information is not exposed. Redaction helps secure user trust, protect personally identifiable Information, and comply with data privacy regulations.
Help link: Redacting Personally Identifiable Information (PII)
Reduce Redundant Efforts with Reusable Functions
Many of your bot elements share similar functionality. Using the Custom Script file, they can now share the code too. Custom Script file considerably reduces redundant development efforts by helping you to collate and upload reusable utility functions. Once you upload a custom script file to the bot, all the functions in the file can be invoked from anywhere in the bot using function calls.
Help link: Reusing Bot Functions with Custom Script File
De-Couple Bot Content from Bot Definitions
Bot building involves extensive content management efforts and coordination between globally-distributed developers and content authors or copyright team. Our newly introduced Content variables make the bot building exercise more seamless by abstracting bot content configuration. Using Content Variables, programmers can avoid hard-coding the bot content into specific bot elements and authors don’t have to be exposed to the bot configurations. Both these parties can rely on language-specific variables file, with the authors building the content into the file and programmers using variables to make content accessible to the bot components.
Help Link: Using Bot Variables
Provided user control over importing and exporting bots
Users can now exercise more control over bot components during imports and exports with new options made available in this release. When you import a bot or a dialog, you can choose one of the following modes to suit your purpose:
- Full import: Completely override the existing bot or dialog with the imported components.
- Incremental import: Select only the relevant components from the imported bot or dialog.
If the import items overlap with the existing bot, the imported bot overrides them. As for exporting, you can export published bots completely or just select specific components for export.
Help link: Importing and Exporting Bots
Support for more languages
Kore.ai bots perform intent detection, custom-named entity recognition, system entity recognition, custom small talk, and built-in small talk in all the following new languages:
- Chinese Simplified
- Chinese Traditional
- Brazilian Portuguese
- Italian
Help link: Building Multi-Language Bots
Added support for Google Assistant channel
We’ve now added Google Assistant as a bot channel. Developers can enable Google Assistant as a channel for mobiles or other devices such as Google Home.
Help link: Adding the Google Assitant Channel
Implemented a new dashboard for universal bots
A new dashboard has been introduced to monitor the performance of Universal bots. The dashboard provides insights on the underlying linked bots as well as the Universal bot.
Optimize trained utterances from ML Model Graph
The Machine Learning Model (ML Model) graph provides an overview of the performance of the trained utterances against the bot tasks. When utterances with sub-optimal performance are identified, you can instantly initiate their training – right from within the graph. In the utterances quadrant, you can reword the utterance to bring the meaning closer to the intent expressed or reassign it to a more-relevant bot task.
Help link: Evaluating Machine Learning Performance
Intent detection in String and Description entities
Many times when you expect users to enter a string or description, they combine their entries with additional follow-up intents. Kore.ai now enables you to configure String or Description entities to capture and respond to an intent using options ranging from ignoring the intent to directly switching to it – with various hold and resume choices in-between.
Help link: Working with the Entity Node
Minor updates
- Simplified task upgrades: Instead of adding individual comments against each task upgrade, the platform now allows you to provide a single comment for any number of task upgrades. Also, the platform lets you track the comments.
- For Batch Testing, in addition to JSON files, you can now also upload CSV files for creating custom suites with the file size not exceeding 2 MB for both CSV and JSON uploads. A visual indicator has been introduced to display the testing progress.
- Two new bot response templates have been added to the FB Messenger channel: FB Login, Quick Reply.
- An option has been introduced to turn on the typing indicator for end users when they are waiting for a bot response.
- New configurations are provided to define methods to extract entities from previous utterances. Help link: Working with the Entity Node
- Bot export no longer prevents developers to switch to other tasks, but runs the process in the background, with the status showing in the same section.