FiscalNote's Machine Learning Demystified

FiscalNote’s analytics are powered by Natural Language Processing and Machine Learning. The former primarily fuels our software’s ability to read and make sense of the language in the bills, rules, and registers that we scrape. The latter of which will facilitate the delivery of more accurate and relevant information based on the user’s actions and preferences.

How will we do this? Well, let’s start with the basics.

Take for example the music streaming application, Pandora—when you select channels and up-vote or down-vote  songs, do you notice that Pandora seems to suggest songs that fall right in line with your musical sensibilities? This is the work of Machine Learning. Pandora gathers your preferences and builds its recommendations off of  the trends in that data. By assessing the patterns in your playlists and songs, Pandora can build rules around your preferences and even begin to predict what music you might want to listen to.

Machine Learning is dependent on your exhibited preferences in order to form an accurate model of your potential preferences. The end goal being to provide you with tailored and relevant results.

So what does this have to do with FiscalNote? Your interaction with legislative and regulatory data is based on the interests and priorities of your organization.

Do you monitor biosimilar rules or manufacturing rules? Do you put a small subset of legislation into watchlists or do you track every bill with the word “tax” in it. With every deleted bill and every watched bill, the platform collects information  about what might be important and what is definitely not worth your time. But we wouldn’t—we couldn’t, really—base any rules or predictions on just a few clicks of your mouse.

Our algorithms require a massive amount of data about your preferences in order to build models that we would trust to deliver you the information that you need.

Enter: the FiscalNote user

With the release of our “Position and Priority” feature, users can now indicate the priority level of bills that they watch, as well as whether they support, oppose, or are neutral to a bill. We have also updated the Discovery Alerts view to allow users to mark a bill as “Irrelevant,” removing it from the Discovery Alert feed entirely.  Not only does this provide more complete analysis and reporting of your legislative priorities, but it also affords FiscalNote the opportunity to learn more about what issues are truly relevant and which to clear out of your workspace.

Machine Learning is not currently being applied to your results—we are in the information gathering stage of the process. Until we have gathered that data, we encourage you to interact with the platform  as much as possible so that we can calculate accurate algorithms and ultimately tailor your FiscalNote experience even further. Removing items from Discovery alerts, adding items to your watchlists, and now setting the priority of bills, are all actions that you can take to ensure that FiscalNote continues to sharpen your policy research—keeping you on top of your legislative game.