We are excited to announce that any Power BI customer can now preview Q&A with their own data! In August we launched Q&A over a set of samples workbooks. Since then we’ve evaluated over 500,000 requests with the Q&A natural language engine. From your feedback, it’s clear that Q&A can fundamentally change how and when data is used: From meeting rooms to casual conversations, you have been impressed with the promise Q&A brings to making decisions faster with data. Since that release, the resounding ask has been to explore your own data with Q&A’s natural language capabilities. We’ve also known that our next update to Q&A would be targeted at enabling usage over custom data. With that in mind, we’ve been working with a set of customers, both internal and external, on using Q&A against custom data models. The feedback from that group has been positive: They’ve found success in using Q&A with their data and find it simple and powerful for identifying and visualizing insights in their data.
Adding your models to Q&A is easy. First you need to upload your workbook to your Power BI Site by clicking on the add link and selecting “Upload file”:
This will upload and enable your workbook for use in the Power BI site. After your workbook has been enabled, click on the three dots on the thumbnail for your workbook and select “Add to Q&A”:
One you’ve completed these steps you can navigate to the Q&A question by clicking the “Search with Power BI Q&A” link in the upper right hand corner of your Power BI site:
That’s it! Once you click on this link, you can begin to ask questions and discover insights in your workbook.
Optimizing your models for Q&A
Once you’ve added your model for Q&A, you can begin to ask questions. The quality of answers that Q&A provides to your questions can be improved by following a set of guidelines and modeling suggestions. Chris Hays, the Program Manager on Q&A responsible for the Natural Language capabilities, has written a detailed set of guidelines on how to optimize your models for Q&A. These model optimizations target four high-level areas:
- Data quality
- Visualization hints
- Synonym modeling
- Clarifying ambiguous questions
Today, you can perform the optimizations described in the first three areas. As we’ve discovered with our early adopters, these three areas enable a great experience with Q&A. One thing we considered was to delay Q&A on your data until we added modeling support for the fourth area. However, based on feedback from our early adopters and to accelerate usage and feedback from all Power BI customers, we wanted to enable all customers to begin using Q&A as quickly as possible.
This release is just the beginning for Q&A on your data. In the coming months we’ll be focusing on improving the system along a number of fronts. We’ll increase the variety of phrases and terms that Q&A innately understands based on your usage of the system, which will directly improve the answers to the questions your users pose to Q&A. We’ll add the ability to augment your models to answer ambiguous questions, addressing the fourth item above. We’ll add reporting on the questions your users are asking with Q&A, enabling you to identify the most popular questions, the ones returning no answers, and a number of other insights. These are just a few of the improvements we’re planning to make in the coming months. We’ll be monitoring the forum and looking for your feedback on Q&A, so go there to post your experiences, questions, and feedback.