One of the most important steps in integrating Brellium is developing good questions. Clients often find disconnects between how our models answer the questions you ask and the way that you would answer those questions yourself. Below are a few tips and tricks to improve the questions which will ultimately make a massive difference in the answers you receive.
Wording your questions can have massive impact on the way they are interpreted by the model.
First, let’s look at some examples of questions that are NOT conducive to good results:
We find when talking with customers that usually these questions are extremely answerable but they just need to be worded a little differently for the model to fully understand them. The name of the game is specificity. Here are a few ways that the above questions could be improved:
For questions that are the most important to you, oftentimes only your clinicians or internal auditors know how to answer them. For these questions, it is important to specify WHERE in the note to find the answer so that our model knows where to look; we like to call this context.
Here are some examples of how to leverage context when asking more complex questions: