The Intent History page in the developer console is a great tool for anyone trying to refine their voice model. The one thing it lacks, when viewing misfiring intents, however, is any indication of the scale of the problem. The output is just a giant list of utterance-to-intent mappings, without any context of how often those failed mappings are happening.

In the case of my skill CompliBot, for example, a recent look at the Intent History output gave me 13000+ unique entries, many of which were hitting the fallback intent when they needed to be hitting actual intents. Searching through 13000 entries and adding all of the thousands of relevant individual values to my intent model is a Sisyphean task, and one that is likely to pollute my intent model in other untestable ways. It wasn't clear to me how deeply I should delve into the list to (and this was making an unproven assumption that the list was actually sorted in frequency order).

As a developer, I would like the ability to see in the Intent History UI the frequency with which a given utterance is mapping to a given intent.

Additionally, I would like some visual queue or numeric indication of how the list is sorting itself.

I understand that there are privacy concerns around how data is provided to the intent history API. It would be perfectly reasonable to obfuscate/generalize this data to not provide specific numbers. Orders of magnitude are what's necessary here. Maybe one reasonable approach would be to label each row with a frequency enumeration from among "HOURLY, DAILY, WEEKLY, VERY INFREQUENTLY" or something.