Power BI User Group of Philadelphia

Please login or click SIGN UP FOR FREE to create your PowerBIUG account to join this user group.

Power BI and Your Organization's Natural Language: A Little Prep for the Dataset

  • 1.  Power BI and Your Organization's Natural Language: A Little Prep for the Dataset

    Top Contributor
    Posted Jul 22, 2020 10:34 PM

    As of 07/21/2020 (Article with Pictures on Linkedin)

    Every organization uses common phrases and has their own natural language which when applied to Power BI data can be extremely insightful. When a Power BI end user is able to ask the Power BI Q&A questions and Power BI is able to answer it correctly then you have empowered the end user with the ability to search for answers and take action. More amazing about Power BI Q&A feature is that the report or dataset designer/architect can edit the linguistic meaning of the dataset.

    This starts with Power BI and Microsoft enabling the designer/architect to add details to the dataset's table or field semantic meaning such as defining the terminology and relationships between entities and attributes. With that, this article explores the required preparation of the model and how the questions asked by the end user exploring the dataset can be updated within the model.

    To begin, lets start with the existing Power BI project, add the Q&A visual to the canvas. This is still found under the visualization pane, but you should use the Ribbon (Insert -> Q&A).

    When the Q&A is added to the canvas, a "Ask a question about your data" will appear. Power BI will also suggest some basic questions that you are able to change however that is not going to be a focus of this article.

    To begin defining the terminology and relationships between entities and attributes, the visual opens a configuration prompt when the settings are accessed on the Q&A visual. Accessing the settings is achieved by selecting the cog wheel icon that is in the upper right corner of the Q&A visual.

    The Getting Started configuration page gives you four main options:

    •  Field Synonyms – Enables adding terms people might use as synonyms for the fields and tables
    • Review Questions - Enables the selection of a Power BI service dataset and review what questions the end users are asking
    • Teach Q&A – You train Power BI the natural-language questions and terms it is not able to recognize
    • Suggested Questions – Suggestive questions for the end user to help them explore the data

    The rest of the article will focus on Review Questions. However, "Suggest questions" allows the designer/architect to suggest possible questions the end user that may be asked to the dataset while the "Teach Q&A" helps Power BI understand some of the basic questions that could be asked.

    The first time that the "Review Questions" is accessed, a dataset that has been previously deployed to the Power BI service must be selected. 

    The dataset that is chosen can be changed at a later point. This is done by following the previous steps to access Review questions, but in the upper right will be an option for "Choose a different dataset".

    First, when end users are asking questions and applying organizational synonyms and acronyms that are not apparent to the data model then those words are highlighted in red with a double underlining score.  In this case, the terminology that has been used will be highlighted red with suggested fields from the underlining (Orange with the term is ambiguous)

    When hovering over the highlighted red word, the end user will be prompted with a message the informs that Power BI was not able to understand what they are asking. More importantly, the end user is prompted with the option to ask the dataset owner to add the terminology to the data set. This feature the question in the Q&A Review Questions in Power BI Desktop.

    When returning to the "Review Questions" section, questions will be listed with the date that the question was asked. Under normal circumstance when no changes are required to fix the terminology, then "Fix requested" and "Fix Needed" both will not have a Pencil icon implying edits required.

    When the Power BI Service identifies that a fix to the terminology will be required then the word is highlighted in red with a pencil marked under the "Fix Needed". If the end user request that a fix is needed, then a pencil will appear under the "Fix requested".

    Selecting on the pencil will navigate to the Teach Q&A pane with the question already populated with the terms that are troublesome identified under the "Define the terms Q&A didn't understand".

    Apply the field or attribute that can be defined in the dataset by typing the value into the "_____ refers to" for each term that the Q&A does not recognize. After applying changes there is more to do. The most important is that the report or dataset must be published, i.e. the Power BI end user of the report won't see this change unless it has been published.

    One last note, the "Manage Terms" section allows any term changes to be deleted. At the current moment, terms are not able to be added or updated under this section. 

    For more information, please refer to



    ------------------------------
    Brian Connelly
    Manager Business Analytics and Insights
    Ashfield Healthcare
    ------------------------------