Show HN: What rapper are you? (my first NLP project in Python/Bottle)
whatrapperareyou.comJust finished this website a few days ago and looking for some feedback.
The program has analysed 300 east and 300 west coast lyrics and looked at the words used in those to train a Naive Bayesian Classifier using the Natural Language Processing Toolkit for Python. It then lets you put in a twitter name and loads its latest 174 tweets and classifies that bag of words to the classifier to let you know if you tweet like west coast or east coast rappers rapped in the nineties.
I finally got NLTK to work online after trying different hosting companies. Decided to learn Bottle to build the web version of the program and it was a breeze, which was awesome as I am really a beginner in programming.
Loved to getting my first real project done and working without quitting before it was finished. Extended my knowledge in jQuery and css a bit and overall had a lot of fun.
I wanted to ask you guys what you think of it and if you have any suggestions both content and design wise.
I found that after getting it to work I was kinda lost in the design. It's funny: when the code works it works, but designing seems to be a never ending job of improving and not being satisfied.
thanks for sharing, i think this is a fun, light-hearted application! i would be interested in seeing a percentage of how east coast rapper i am (60% east coast / 40% west coast) just to see the break out. also, it would be cool to show some of my tweets that feature some of those keywords, to remind me (as i couldn't recall any tweets that had those words).
aside from that, great work!
thanks for the comment, means a lot! Good suggestion as well, the percentage one might be too hard as I'm using some basic NLTK libraries, but the one showing tweets I'll definitely look into!
Furthermore, the words listed are certainly not necessarily words you used, but just the 50 most differentiating words between east and west coast. 'compton' is used way more in west coast lyrics than east for example. It's just there to see some fun differences not related to your personal twitter stream. At last, in total the classifier looks for not 50, but 1000 words in classifying your tweets. Hope this explained it a bit.
It's telling me "That ain't no real twitter user bro!" which is false (and kind of lame).
I'm sorry to hear that. When Tweepy gives an error I let it output that sentence as almost every time it was a Twitteruser does not exist error. I checked your twitter and its working with me so it might be because of some exceeded twitter daily call limit or something on your network. I noticed that when I was developing it -- after a day of refreshing and reiterating it would give that error too.
Thanks for pointing out the flaw, I'll try to implement some more descriptive error messages for different errors.