I was testing Watson personality insights API as a part of my application testing and found that when I tested with an input text consisting of just number, Watson PI is returning proper summary.
I would like to inform it to the Watson Personality Insights Dev team, as this bug may make people doubt Watson's ability to understand the input.
You have a good point, in that it indeed looks bad, even though it is technically is a correct response.
One of the (many) ways we categorize the words, is numeric. So, having text that only has numbers, is just as valid as text which for instance only has angry words, even though both will provide a skewed profile. While technically correct, the service should be checking the input text to see if it is highly concentrated in a single type of word, and if so either provide an error or a warning.
We will continue to provide updated on the issue you opened at https://github.com/watson-developer-cloud/personality-insights-nodejs/issues/89