Product Update - v5.5, Webex Connect Bot Builder Platform, July 2023

Webex Connect Bot Builder v5.5 release brings multiple new capabilities to help you continue delivering seamless customer experiences. Here’s a list of the key updates:

• Analytics revamp
• On-demand report generation
• Context in Q&A bots
• Article editor revamp
• Option to include partial match in testing results
• Ability to filter testing results
• NLP pipeline transactions will be encrypted by default
• Inclusion of active context information in sessions
• Addition of timeout responses in voice channels

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

Changed - Analytics revamp

The Analytics section helps developers evaluate the bot performance and its effectiveness by looking at key metrics. In this release, the analytics section is revamped to introduce new metrics and visualizations that are more useful for bot developers to interpret their bot’s usage as well as its training data health and curation state.

The analytics are now divided into four sections represented as tabs. These are: Overview, Responses, Training, and Curation. On visiting the analytics screen, developers can select the bot they want to see analytics for. They can also customize the analytics view by choosing the channel they want to see the data for, along with date range and the granularity of the data. By default, analytics data for the last month is shown for all channels with a daily granularity (each day being a point on the x-axis in the graphs).

Overview

Overview contains key metrics and graphs that provide a snapshot of overall bot usage and performance to the developers. The first section in the overview displays the following statistics about sessions and messages for the bot:

• Total sessions and sessions handled by the bot without human intervention.
• Total agent handovers, which is a count of the number of sessions handed over to human agents.
• Daily average sessions
• Total messages (human and bot messages) and how many of those messages came from users.
• Daily average messages

This is followed by a graphical representation of total sessions and total messages.

The second section contains stats about users for the bot. It provides a count of total users and information about average sessions per user and daily average users. This is followed by a graph displaying new and returning users for each unit depending on the selected granularity.

The third section provides stats about bot’s responses to users. Here one can see total responses sent out by the bot and the split up between responses where the bot:

• Identified the user’s intent.
• Responded with a fallback message.
• Responded with a partial match message.
• Informed the user of an agent handover.

The same is aggregated in a pie chart and an area graph provides information based on selected granularity.

Responses

Give the developers a detailed view of what the users are asking about and how often they are asking it. It provides a graphical representation of the most popular articles for Q&A bots and response templates for Task bots.

This section also contains a tabular breakdown of the above. The table contains:

• Article/Template key
• Category/Intent
• Count – Number of times an article/intent was invoked in the selected time period
• Conversations – Number of conversations an article/intent was invoked in for the time period
• Avg. confidence – Average confidence score with which the article/intent was detected across all invocations

Training

The training section of analytics is representative of the ‘health’ of a bot corpus. It is recommended that developers configure 20+ training utterances for each intent/article in their bot. In this section, all the articles/intents in a corpus are displayed as individual rectangles where the color and relative size of each rectangle is indicative of the training data the article/intent contains. The closer an intent is to white, the more training data it needs for your bot’s accuracy to improve.

Curation

Provides a visual summary of how many curation issues have been coming up each day and how many of them have been resolved by the bot developers.

Changed - On-demand report generation

In the previous versions of bot builder, enterprise users could configure various kinds of reports and get them delivered via SFTP, email or to their S3 bucket. The configuration required them to schedule the reports that would be delivered to them on future dates after fixed intervals. With this release, users will be able to generate reports for specified periods on-demand and get them delivered via existing delivery mechanisms. The maximum duration for which a report can be generated is 30 days.

These reports will also be available for download from the history tab in Reports.

Added - Context in Q&A bots

Natural language conversations we have in our everyday life are made easier and more succinct by the presence of context. Things like ‘I want to buy that’ are easily communicated as long as there’s enough context to understand what ‘that’ is referring to. We are introducing contexts in Q&A bots to handle conversations like these with consumers. Expressions like these can be matched with an article if the necessary context is provided.

Entry and exit contexts for an article can be configured within the article editor section.

To understand context in Q&A bots better let us go through an example. Consider a bot that has been configured for the job portal for an online delivery company. One article speaks about the process, another speaks about the openings and there is one more article about the pay scale. If a user configures an exit context called ‘Jobs’ in the first article and keeps it the same as the entry context in the second, ‘jobs’ will become the exit context for the bot conversation for the next 5 turns and every time a user asks about the ‘delivery job’ the bot will send the appropriate response while retaining the ‘delivery’ context.

Exit context

Exit contexts control the active contexts for a session. An exit context contains the context value string and the duration of that context. When an intent is completed, the configured exit contexts for that intent become exit for their respective durations. Developers can configure a maximum of 15 exit contexts for a particular article. An exit context can be added by pressing the enter/return key after typing the context.

Entry context

Entry contexts control whether an article can be matched with the end-user query based on the active context of the session. When context is present in a session, the following rules are applied for article matching: 

• An article with entry contexts will only be matched if the active context in the session already contains all the required entry context values. In other words, the entry context of an article must be a subset of active context for it to be matched. 
• For all intents satisfying the above rule, preference will be given to articles whose input context matches the active more closely if the confidence scores for multiple intents are the same. In other words, input context will be used for tie-breaking partial matches. 

Developers can configure a maximum of 5 entry contexts for a particular article.

Changed - Article editor revamp

The information architecture for the article editor page has been updated to make the overall user experience better. Tabular distribution of languages makes it easier for the user to surf between different languages. The article name and category fields have also been arranged in a more accessible fashion with an even simplified user experience.

Added - Option to include partial match in Testing results

The results of test case executions contained information about how many test cases were pending or had passed/failed based on the expected article/template. All the test-cases that resulted in partial matches were considered as failed test cases even if the expected response was presented to the users in partial match options. This is undesirable in a lot of cases where configured intents are similar to each other and providing options to end-users to select the appropriate intent is not a drawback. To solve this, a new ‘Include partial matches’ toggle is added to testing sections which determines whether partial match results that include the expected result will be considered as passed or failed. ‘Passed with partial match’ is also included as an additional outcome in the execution tab of Testing.

Added - Ability to filter Testing results

Testing results in the execution tab can now be filtered by clicking on the desired result. Users can now look for test cases with a particular result without having to scroll through all the executed issues.

Changed - NLP Pipeline transactions encrypted by default

Similar to Q&A and Task bot sessions, NLP Pipeline transactions are not encrypted by default and can only be accessed by users with ‘Decrypt access’ permission. Any test transactions done from the bot builder UI are considered as test transactions and are decrypted by default. All other transactions require decrypt access user permission.

Added - Active context info included in sessions

Transaction info in sessions now contains the state of context for each transaction. Developers can now see what the active context was at any point in the conversation between the bot and the user.

Added - Timeout response for Voice channel

In voice settings for each response for the voice channel, users can now configure a ‘timeout response’ along with timeout duration. This response will be played back to the user if they do not respond to the message within the duration specified in the ‘Timeout (s)’ field. The orchestration for voice channel is done via WxCC voice flow builder (limited alpha for now).

Changelog