Quick start for Topic Recommendations

  • Release version: Xanadu
  • Updated August 1, 2024
  • 2 minutes to read
    • Summarize
    Summarized using AI
    This content was generated using new OpenAI-powered functionality. Results are provided on an as is basis and are not guaranteed to be accurate or complete.

    Summary of Quick start for Topic Recommendations

    The Topic Recommendations feature in ServiceNow Virtual Agent helps you identify and add relevant conversation topics based on your existing data. By configuring and running analysis reports, you receive recommended topics that reflect frequent user intents in your data sources, enabling you to enhance your Virtual Agent's conversational capabilities quickly and effectively.

    Show full answer Show less

    This functionality is accessible via the Topic Recommendations page located at All > Conversational Interfaces > Virtual Agent > Topic Recommendations in your ServiceNow instance.

    Key Features

    • Configurable Analysis Reports: Define reports based on various data sources such as Incident, Request tables, Live Agent chat transcripts, and Virtual Agent fallback utterances. You select specific string fields (e.g., Short description) to analyze and can apply conditions and scheduling for automated report runs.
    • Intent Discovery Mapping: The system compares your data fields against a taxonomy of business applications to discover intents. It then maps these intents to pre-built Virtual Agent topics or identifies new topic recommendations when no existing topics match.
    • Topic Recommendations Display: Analysis results appear as topic cards sorted by frequency of matched intents, showing the top 20 intent matches with the option to explore additional intents in the NLU Workbench.

    Using the Feature

    • Step 1 - Configure Analysis Reports: Set up reports in Virtual Agent settings to specify data sources, fields, and taxonomies. This step populates the Topic Recommendations page with actionable data.
    • Step 2 - Run Reports: Execute all or selected reports from the Topic Recommendations page. The platform performs intent discovery and topic mapping to produce recommendations.
    • Step 3 - Review and Act: Review categorized results: topics Virtual Agent can already handle with pre-built topics, and user intents requiring new topic flows. You can add recommended topics by duplicating pre-built ones for customization or create new topics linked to matched intents for flow development in Virtual Agent Designer. Existing topics can also be linked to recommendations.

    Benefits for ServiceNow Customers

    • Quickly identify relevant conversational topics driven by actual user data and intents.
    • Leverage pre-built topics to accelerate Virtual Agent topic deployment.
    • Expand Virtual Agent capabilities by creating new topics where needed, ensuring comprehensive coverage of user requests.
    • Streamline topic management with clear insights into data-driven recommendations and easy integration into Virtual Agent Designer.

    To get recommended topics based on your data, configure analysis reports, run the reports, and then review the results. You can either add new topics to Virtual Agent or link a recommendation to an existing topic.

    The Topic Recommendations page appears in your instance at All > Conversational Interfaces > Virtual Agent > Topic Recommendations.

    Step 1: Configure analysis reports based on ServiceNow data

    Until you define analysis reports, the Topic Recommendations page is empty.

    Figure 1. Topic Recommendations page with no reports defined
    The Topic Recommendations page with no analysis reports defined.
    Each analysis report is based on the following:
    • Data source: The ServiceNow data that you want to analyze. The options are: Incident [incident] table, Request [sc_request] table, Live Agent chat transcripts, and Virtual Agent fallback utterances.
    • The field to analyze: A selected field (with a string data type) in the data source table to analyze. For example, the Short description field.
    • Taxonomy: The business application that includes the data sets. When you install ITSM Virtual Agent Conversations, the Incident and Request data sets are installed by default. As additional data sets for applications become available, the data sets are displayed as a taxonomy on the page after they’re installed on an instance.

    You can also add conditions to your analysis reports and configure them to run on a schedule. When new recommendations are available, a notification appears in the Virtual Agent Designer Topics page.

    Configure analysis reports for Topic Recommendations in Virtual Agent settings. For details, see Configure analysis reports for Topic Recommendations.

    Step 2: Run an analysis report

    Run all configured analysis reports together or selectively run one at a time from the Topic Recommendations page.

    During analysis, the Topic Recommendations app does the following:
    • Uses Intent Discovery to review your data source, comparing your data source against the related taxonomy for your business application. Intent Discovery matches the data field specified in your data set to intents in the taxonomy.
    • Maps the matched intents to pre-built topics in Virtual Agent. If intents don’t have associated pre-built topics, the app recommends them as new topics that you can build because they were requested frequently enough by your users.
    • Displays the analysis results as topic cards in the Topic Recommendations page.

    For more information, see Run a Topic Recommendation analysis.

    Step 3: Review report results and add or link topics and intents

    Results are sorted into the following groups:
    • Topics that Virtual Agent can handle: Recommendations in this group can be addressed with a pre-built topic.

      When you add this type of recommendation to Virtual Agent, the matched pre-built topic is duplicated in your instance. You can then modify, test, and publish the added topic using Virtual Agent Designer.

    • User intents that need topic flows: Recommendations in this group match a ServiceNow NLU intent.

      When you add this type of recommendation to Virtual Agent, an empty topic is created and mapped to the matched intent in the model that you specify. You can then open the topic in Virtual Agent Designer and create a flow for the new topic.

    The results are displayed in descending order of matches to your data source. The top 20 intent matches are displayed. If there are more than 20 intents, you can view them in NLU Workbench.

    For more information about adding a recommendation to Virtual Agent, see Add recommended topics and intents to Virtual Agent. You can also link a recommendation to an existing topic.

    Figure 2. Mapping recommendations to Virtual Agent topics
    Topic Recommendations analyzes your data and combines it with an understanding of the product area to suggest pre-built topics and intents or new content for Virtual Agent.