There are times when you want to see what trends show up in your data and expose the unknown unknowns. But other times you just need to know the answer to a question:
“Is our new work from home policy causing our engineers to feel unappreciated?”
Or if anyone is mentioning a critical topic:
“Does anyone feel like they are in danger at work?”
Everyone has their own list of hypotheses or critical topics, and we can detect if anyone mentioned them in their responses. Even a single response on the topic can be critical to identify so immediate action can be taken.
Here are the goals of this system:
A single response can be critical
Identify all responses related to the topics, even if there is only 1
Change your mind about what is interesting
Dynamically define the topic or topics of interest in real-time
Change your mind about how similar responses need to be to the topic
Dynamically adjust how close to the specific topic the responses must be
We accomplish this by first leveraging our unbiased topic generation analysis that calculated the semantic meaning of every response. Then in real time, we use the same technique on the user provided topics and ask our clustering algorithms to identify the responses that have the same or similar meaning to each user defined topic.
The result is a real-time search that is based on the semantic meaning of a topic instead of the keywords you choose. This allows you to confidently determine if anyone mentioned a given topic, not trying to guess what words they might have used to describe that topic.