This online demo presents SemScale, a tool for semantic scaling of political text, originally described in:
Glavaš, Goran, Federico Nanni, and Simone Paolo Ponzetto. "Unsupervised Cross Lingual Scaling of Political Texts." Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics, 2017.
The demo has been developed by Ferdinand Betz under the supervision of Federico Nanni, and is quite simple to use:
- zip together the .txt documents you want to scale and upload the .zip file (at least 4 .txt files)
- (optional) upload a .txt file with a list of stop words (one per line)
- select the language of the uploaded documents (we currently support English, German, French, Italian and Spanish)
- click submit! (It will take a few minutes, depending on the size of your documents)
The demo relies upon general purpose FastText embeddings and currently uses the 100k most frequent words for each language under study, for performance reasons.
For research purposes, we recommend you to use the Python implementation of SemScale, which is available here.
For further information about the online interface contact: firstname.lastname@example.org