Product Update - v5.7, Webex Connect Bot Builder Platform, February 2024

Webex Connect Bot Builder v5.7 release brings multiple new capabilities to help you continue delivering seamless customer experiences. Here’s a list of the key updates:

• Support added for GBM (Google Business Messages) as a channel.
• Introduced training data generation feature for task bots to reduce the initial build time for bots.
• Added support for entity roles for system entities for Mindmeld NLU engine to make certain use cases more conversational.
• Code generation and Text summarization NLP pipelines for usage and tracking.
• Changes to advanced engine settings behavior for multi-lingual Q&A bots.
• Added contextual redirection for users cross-launching from the control hub.
• Link to article experience modifications in quick replies template.
• Streamlined Intent editor user experience.
• Streamlined Enterprise Profile UI by adding new information and getting rid of obsolete fields.
• Fixed a bug where admin users were not able to give bot access to non-admin users.

Please refer to the details below to learn about all the changes and enhancements.

📘

Release Date

The date of release will be communicated separately over an email.

Details

Added - GBM (Google Business Messages) as a channel

The platform now supports a new channel – Google Business Messages. Now, businesses can effortlessly incorporate Google Business Messages into their conversational strategies. At present we support the following code templates -

• Text
• Reply suggestions
• Code response templates

Added - Training data generation feature for task bots to reduce the initial build time for bots

Training data generation capability focuses on the “cold start” problem with bot building - developers having to manually add training data to their intents to get the bot working at a reasonable accuracy. The training data consists of different ways in which a user can invoke the same intent. We recommend that users add at least 15-20 variants for each intent to improve bot accuracy. Creating this training corpus manually can be quite tedious and time consuming. In practice, we have observed that developers tend to cut corners in this crucial part of bot building by adding only a few variants, or by adding only keywords as variants instead of meaningful sentences. This new feature provides users with Large Language Model (LLM)-powered training data set generation, which will allow chatbots to be created rapidly with significantly less initial time investment. Developers can supplement each intent’s training data by clicking on the ‘Generate’ button after providing the intent name and adding one sample utterance. This will open a dialog box where users can provide a brief description of the intent. This description plays a crucial role in the quality of variants generated. Developers can also select the number of variants they would like to generate and how creative they would like the underlying LLM
to be while generating these variants.

Added - Support for entity roles for system entities for Mindmeld NLU engine to make certain use cases more conversational

Developers can now use entity roles for system entities for the Mindmeld NLU engine. As a prerequisite for using this, entity roles need to be enabled in the Advanced settings section of the Change training engine dialog box. Similar to custom entities with roles, a single system entity can now have multiple roles. This is especially useful for use cases where the same
entity must be collected multiple times to complete an intent. For instance, you want to book a hotel from date A to date B. A single date entity can be created with ‘from’ and ‘to roles. Subsequently, annotate the training utterances in the intent with these entity roles. The roles arranged in a specific pattern will allow the chatbot to learn the pattern in consumer utterances and have a seamless conversation with the consumer.

Changed - Advanced engine settings for multi-lingual bots

For better alignment with the advanced engine settings viz-a-viz vectorizer in multilingual bots, we have made some modifications to the conditions. Here are the updates:

• Remove stop words, wordform expansion, synonym, and wordforms checkboxes will be disabled when English isn’t there in bot’s languages
• Keep ‘remove special characters’ as a selectable option in all cases

Added - Control Hub users to be redirected to the relevant path in Bot Builder

Developers can cross-launch bot builder from Control Hub to create task bots for voice channel. When bot builder is launched from Control Hub, users will be redirected to the Task bots page instead of the Q&A bots page as the Limited Availability release of bot builder for voice flows only supports task bots. Further, when launching the bot builder from the voice flow builder canvas, developers will be redirected to the bot they are using in their voice flow instead of the dashboard.

Changed - Link article to payload UX

While searching for an article in the link to payload experience, earlier both articles and utterances used to be displayed as search results. Modifications have been made to the search results by displaying only article names. There are cosmetic changes to the search results as well.

Changed - Changes to Enterprise Profile

Enterprise profile section now contains a tenant identifiers section (read-only) that contains the Webex org ID, CPaaS org ID and Subscription ID for the tenants where it’s available. Further, some deprecated fields have been removed from the Profile Settings section to streamline the UI.

Changelog