Klemens, could you explain exacter what you mean? What is a well--trained model for sentiment analysis in English and where I can find it? What corpus did Stanford use for training etc?

Alexander,

with the Stanford Core NLP framework, sentiment analysis can be carried out. To determine if a sentiment is positive or negative, the Stanford Core NLP Framework uses domain-specific sentiment lexicons induced from unlabelded corpora (see William L Hamilton, Kevin Clark, Jure Leskovec and Dan Jurafsky. 2016: Inducing Domain-Specific Sentiment Lexicons from Unlabeled Corpora). Unfortunately only Corpora in English language are used for the lexicons. Hence my question: Are there already sentiment lexicons that have been "trained" with German corpora?

I didn't work with German lexicons since my PhD was in opinion mining and lexical affect sensing in English Thesis Opinion Mining and Lexical Affect Sensing

- German sentiment lexicons can be built on the basis of the NLTK corpora https://www.nltk.org/book/ch02.html

- German Wordnet can be used to build a sentiment lexicon https://mailman.uib.no/public/corpora/2018-April/028257.html

- There is also the German LIWC that contains sentiment words and can be used for your purposes https://s3-us-west-2.amazonaws.com/downloads.liwc.net/LIWC2015_OperatorManual.pdf

Anyhow, in most cases you must consider translating from English in German, for example, if you want to use Jurafsky's approach in German.

Thank you!

I recommend Alexander's answer

Best Regards