I recently completed a seven-post series on the IBM Watson Developer Blog about Cognitive System Testing. These posts reflect some hard-learned lessons from the past few years. Cognitive systems present some interesting challenges for testing-minded people, some of these challenges are unique to cognitive systems and some are common to all software systems.
Part 1: How to test a cognitive system (and why it’s so important), this includes guiding principles and sets up the series.
Part 2: Cognitive system testing: Smoke testing, after all the most important question to answer quickly is “is my cognitive system on fire?”
Part 3: Cognitive system testing : Testing at the beginning with ingestion verification test, since a cognitive system is only as good as the data you are loading into it.
Part 4: Cognitive system testing: Natural language processing unit testing, because testing NLP is both hard and important.
Part 5: Cognitive system testing: Overall system accuracy testing, users judge the system by its accuracy and you should too.
Part 6: Cognitive system testing: So, I don’t need to unit test my cognitive system?, I hope you weren’t going to forget about unit testing!
Part 7: Cognitive system testing: Concluding remarks, wrapping up the series with reminders about where to start and stop your testing process.
I hope you find these posts useful, please leave a comment if they are!