Mystery Friend

RETRIEVAL-BASED CHATBOTS

Mystery Friend

Hi,

I just finished with the Mystery Friend project. Is anyone can take a look to the code please…
I modified the original code to increase readability. All test text inserted to the modules. You can comment out any unwanted test and easy to select another one.
In the script.py / main.py file from line 19 to line 32 you can modify the test data:

  • Postcard test
  • Emma Goldman test
  • Matthew Henson test
  • TingFang Wu test
  • an email test

Everything works well (maybe) but I need a little help how to improve the code because when I run the email test, I get Emma Goldman as result with predict_proba: 98%. However is false.
Maybe I need to create another test for .score(test_vectors, test_labels) or any suggestion?

Link to GitHub
Link to Repl

Csaba

I’m sorry but I haven’t done this project as this was not on my career path or any other courses that I have taken. But I know python, so if you could brief me about the project I might be able to help you out with the code review.