Step 3: Refine Your Keywords
Keywords are the lifeblood of most policy monitoring systems. And since they're a product of language, which is always changing, they need to be updated often.
In politics, new buzzwords can pop up overnight. It's also common for a keyword you've been tracking for one topic to suddenly get associated with a completely different topic.
Take the word "hallucinations." Prior to 2023, the word was mostly used to talk about mind altering drugs or medical conditions. Today, the word is often used when general purpose large language models (like ChatGPT) produce erroneous hallucinations.
To make sure you're monitoring the right keywords, take a look at the keywords that triggered your most relevant alerts as well as your most irrelevant. You can do this by using the built-in relevancy scoring algorithm.
To find keywords that are generating more noise than signal, you can reverse the sort to surface the most irrelevant alerts. Then, click "View Details" to see which keywords triggered those alerts. Those keywords might need to be refined with additional words, or the alert might need to be deleted if the term has been co-opted by an unrelated conversation.
PolicyNote has another feature that can help you surface relevant policies when narrowly defined keywords don't exist. The AI-Assisted Search bar lets you type natural language queries — such as "Medicare bills that impact pharmacy benefit managers" — to surface existing policies that match not only the phrase you typed, but also thematically related concepts.