AFINN sentiment analysis in Python: Wordlist-based approach for sentiment analysis.
>>> from afinn import Afinn >>> afinn = Afinn() >>> afinn.score('This is utterly excellent!') 3.0
In Danish:
>>> afinn = Afinn(language='da') >>> afinn.score('Hvis ikke det er det mest afskyelige flueknepperi...') -6.0
In Finnish:
>>> afinn = Afinn(language='fi') >>> afinn.score('Siellä on uusi hyvä juttu, katsokaa ja kuunnelkaa ihmeessä.') 3.0
In Swedish:
>>> afinn = Afinn(language='sv') >>> afinn.score('det är inte bra') -2.0
In Turkish:
>>> afinn = Afinn(language='tr') >>> from six import u >>> afinn.score(u('iyi de\u011Fil')) -2.0>>> afinn = Afinn(language='tr') >>> afinn.score('iyi değil') -2.0
With emoticons:
>>> afinn = Afinn(emoticons=True) >>> afinn.score('I saw that yesterday :)') 2.0
With multiple sentences (here with data from an Austen novel available in Gutenberg):
>>> from afinn import Afinn >>> from nltk.corpus import gutenberg >>> import textwrap >>> afinn = Afinn() >>> sentences = (" ".join(wordlist) for wordlist in gutenberg.sents('austen-sense.txt')) >>> scored_sentences = ((afinn.score(sent), sent) for sent in sentences) >>> sorted_sentences = sorted(scored_sentences) >>> print("\n".join(textwrap.wrap(sorted_sentences[0][1], 70))) To attach myself to your sister , therefore , was not a thing to be thought of ;-- and with a meanness , selfishness , cruelty -- which no indignant , no contemptuous look , even of yours , Miss Dashwood , can ever reprobate too much -- I was acting in this manner , trying to engage her regard , without a thought of returning it .-- But one thing may be said for me : even in that horrid state of selfish vanity , I did not know the extent of the injury I meditated , because I did not THEN know what it was to love .
If you as a scientist use the wordlist or the code please cite this one:
- Finn Årup Nielsen, "A new ANEW: evaluation of a word list for sentiment analysis in microblogs", Proceedings of the ESWC2011 Workshop on 'Making Sense of Microposts': Big things come in small packages. Volume 718 in CEUR Workshop Proceedings: 93-98. 2011 May. Matthew Rowe, Milan Stankovic, Aba-Sah Dadzie, Mariann Hardey (editors)
Paper with supplement: http://www2.imm.dtu.dk/pubdb/views/edoc_download.php/6006/pdf/imm6006.pdf
- http://neuro.compute.dtu.dk/wiki/AFINN - Brede Wiki entry on AFINN with pointers to many scientific papers
- https://github.com/darenr/afinn - Sentiment analysis in Javascript with AFINN word list