As we can see, the strategy explained previously sounds very much like usual tests. One might say, why can't we use unit tests instead of implementing another component/service for measuring accuracy? The answer is because our intention is not to test the application in order to find flaws. The point is, tests are a way to find flaws in what we EXPECT from a piece of code. Take a look at the following test, for example:
it('should test TF-IDF accuracy', () => {
expect(/* query outcome */).toBe(/* valid */);
});
Note
The keywords such as "it", "expect", and "toBe" are the real function names used in testing. If you are interested in learning more about testing try: Jasmin Cookbook (https://www.packtpub.com/web-development/jasmine-cookbook).
So this code snippet here says we want 'to test TF-IDF accuracy' and we 'expect that our Custom Search Engine results are valid'. And how possibly can we guarantee that! As you might have guessed, there...