When I asked for advice on how to write a test, I was asked where the test was located, and when I tried to answer that it was the same as an existing test, I realized that it didn't have to be the same. https://keicho.netlify.app/#talk=hWLZQImnEQxmsWK4WDC8
What would you like to see happen in this conversation?
I'm thinking of making a slightly large-impact modification to this chat system, and I think it will probably work, but I'm wondering how to write a test, and I'd like to discuss it with you.
What type of chat system is that?
Repeat questions for keywords and dig deeper.
What type of keywords are those?
It's tied to an important concept in the user's mind.
What type of user is this user?
NGKW Users
What type of concept is that concept?
Still ambiguous
I'm thinking of making a modification to this chat system that will have a slightly larger impact, and I think it will probably work, but I'm wondering how to write a test for it, so I'd like to discuss it.
I would like to figure out how to test this. Questions about the relationship between keywords should be asked after each keyword has been drilled down, but currently, when a keyword with a very high attention score appears, the relationship between it and a word that has not been drilled down is asked.
Is there anything else you can tell us about those keywords?
Newly appearing words have a low attention score. Repeated occurrences will be higher. Should be developed by digging deeper before asking relationship questions.
What type of word is the word?
Same as keyword
Where are those keywords?
It is in the user's text.
Same as keyword
I'm more interested in discussing testing methods.
What type of test is that?
I want a test that can confirm what kind of bad situation we're in, and that the fixes have improved it, and automatically confirm that this case hasn't gotten worse when we tweak things in the future.
Is there anything else you can tell us about that test?
It is difficult to write a test because it is not the kind of test that says, "When there is an input like this, react like this," and I don't know how to write it.
Where is that test located?
I was going to put it with the other obvious tests... but maybe that's a mistake? Maybe I should stop making implicit assumptions about the test runner I'm using now?
Where is that test located?
You can either cut another folder and put the non-question-by-question test in there.
How do you know that test?
What I wanted to test was "no questions asking about relationships before digging questions" or "no bias towards one question type or one keyword", so I wonder if I could pour user input from the logs to test the former kind of back-and-forth conditions, or get statistics for the latter?
What is it that you value?
Testing increases the speed of product growth by providing peace of mind when making modifications that have a large scope of impact.
What does that test look like?
Like a helmet or knee pads to wear when practicing roller skating. Reduces the risk of injury.
Testing increases the speed of product growth by providing peace of mind when making modifications that have a large impact scope.
It is good to be able to describe in the above way what is difficult to describe in each question.
What needs to happen for that to test?
If the final test returns a true/false value, it can be integrated with conventional tests, but there are still some that do not, so it is necessary to separate them from the current tests once they are in place
Can that test do it?
Then we'll do it tomorrow.
If it eventually returns a true/false value, it can be integrated with conventional tests, but some still do not, so it is necessary to separate it from current tests for once
This page is auto-translated from /nishio/会話ログ2021-01-31 using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I'm very happy to spread my thought to non-Japanese readers.