We're going to to perform some tasks concerning sentiment classification of tweets.
Which lexicons, dictionaries, and other types of corpora do you think should be taken into account?
Could you give links to such suitable corpora?
Newspapers and magazines often classify their articles under names like politics, finance, business, sports, science, technology, health, etc. Collections of articles like that under specific classes may form the best corpora for your purpose. Tweets tend to be rooted in our times, deal most of the time with current affairs, and are often triggered by news reports.
This paper recently introduced an unbiased general purpose dataset for sentiment analysis of Tweets which is publicly available: https://www.researchgate.net/publication/282148966_Sentiment_Uncertainty_and_Spam_in_Twitter_Streams_and_Its_Implications_for_General_Purpose_Realtime_Sentiment_Analysis
Link to the dataset: http://project2.cs.uos.de/TweeDOS/
Conference Paper Sentiment Uncertainty and Spam in Twitter Streams and Its Im...
I suggest you take a look at the attached link.
You will find many lexicons and usefull corpora for English sentiment classification.
Personally I used the EmoLex and it helped me to improve my accuaracy, but you can use it along with the others: The Hashtag Emotion lexicon and Corpus, Yelp and Amazon Sentiment Lexicons...
http://saifmohammad.com/WebPages/lexicons.html
If you want to work with French Data. We translated and expanded the EmoLex to French. Please check the attached link.
http://www.lirmm.fr/~abdaoui/FEEL.html
I think you should follow the dataset of Socher at Stanford and also follow other dataset of SemEval.