aboSamoor / polyglot

Multilingual text (NLP) processing toolkit
http://polyglot-nlp.com
Other
2.31k stars 337 forks source link

Editing sentiment models #203

Open Scilent0r opened 5 years ago

Scilent0r commented 5 years ago

Hey,

Are there any possibilities to edit the sentiment models and how is it done?

After downloading the sentiment2.fi -file and checking what the .pkl contains, the contents are random words in a list without annotation of pre-determined 'positive' or 'negative' values as pasted below:

Partial content from finnish sentiment pickle (end): _

'älyllinen', 'äpärä', 'ärsyttävä', 'ärsyttää', 'äänekäs', 'äänetön', 'ääniä', 'äärettömän', 'äärimmäisen', 'šokki'] 1 : [[-1] [-1] [-1] ... [-1] [ 1] [-1]]"

_

Is this the part that pre-determines the sentiment of a word, and are there any best practices on editing them (if its done in this file) to improve the accuracy?

shamooosa commented 3 years ago

+1 .. did you found the answer?