search

bnosac / pattern.nlp

R package to perform sentiment analysis and Parts of Speech tagging for Dutch/French/English/German/Spanish/Italian
Other
67 stars 17 forks source link

Dutch sentiment analysis not working #12

Closed frederatic closed 5 years ago

frederatic commented 5 years ago

I got it working, but when I try your examples for pattern_sentiment, the dutch one outputs 0.00 for both polarity and subjectivity. I tried the same example in python 2.7 with pattern 2.6 and it gives the same results, so I assume the problem lies with pattern. Problem is I can't install the latest pattern version with python 2.7, cuz it says it requires >3.5. And this R package can't use 3.5 as it says in your Read.me. How do I fix this so I can get the same output as yours?

jwijffels commented 5 years ago

Try a bit more :) on installing pattern in python. https://www.clips.uantwerpen.be/pattern

frederatic commented 5 years ago

Only problem I can see from that page is that I installed both pattern 2.6 and 3.6 using setup.py install since pip install gave errors. I tried their examples in python 2.7 (pattern 2.6), python 3.6 (pattern 3.6) and your package in R.studio: all gave the same results matching their output. For your example, it gives 0.00 in both python 2.7 and R, but (-0.2, 0.4) in python 3.6. I have no idea how this can happen, so can you help me out? I have to make sure the sentiment analysis is as accurate as possible once I apply it to a large data set.

frederatic commented 5 years ago

Their example: "Een onwijs spannend goed boek!" Your example: "We waren bijna bij de kooien toen er van boven een hoeragejuich losbrak alsof Rudi Vuller door Koeman in z'n kloten was geschopt."

jwijffels commented 5 years ago

I don't understand your problem. It seems to be working? This R package executes py <- findpython::can_find_python_cmd(required_modules = "pattern.db") at startup of the package to connect to pattern. If it finds that one it will use it. What does that command give for your computer.

frederatic commented 5 years ago

Yes, it works, but the dutch example in your Read.me outputs different results for me (see above), so that worries me it won't be accurate on other sentences.

jwijffels commented 5 years ago

Why don't you just remove either the python2 or python3 pattern module, whichever you don't like.

frederatic commented 5 years ago

Tried, did not change any results. It works fine on other sentences, so I think that specific example is just an outlier. Can you try the examples posted above (both in python and R) and send me your output just to make sure?