Lexical sentiment analysis made with Streamlit [Leksykalna analiza emocji w tekście]
The app allows to analyse sentiment and emotions in text with the use of a lexicon-based method. The app provides an analysis for Polish and English language. In regard to Polish, the analysis can be conducted with 1 out of the 3 available lexicons (Emotion Meanings, NAWL, combined Emotion Meanings and NAWL). In regard to English, the analysis is conducted with the NRC-EMOLEX. The app is developed with the use of Streamlit library as part of collaboration in the ComPathos project.
Argument Interchange Format (AIF) Converter made with Streamlit
The app allows to analyse the annotation of Inference Anchoring Theory (IAT, Budzyńska & Reed 2011) encoded in AIF format (see more: AIFdb guide) - relations between arguments, speakers' turn taking, as well as conversion of .json format to .csv. The end user can upload so-called maps (nodesets) in .json format or provide an id of the map (nodeset) stored directly in AIF database. The app is developed based on the code provided by roryduthie.
RoBERTa-large stance classifier [Stance-Tw]
The model allows to recognise a stance towards an entity mentioned in text (attack, support, neutral). Developed on a sample of over 3k sentences from Reddit and Twitter posts, and manually annotated with stance. Model developed in a collaboration with Laboratory of The New Ethos.
XLM-RoBERTa-base sentiment classifier [sentimenTw-political]
The model allows to detect 3 categories of sentiment (negative, neutral, positive) in text. Developed for multilingual analysis with the XLM-roBERTa-base model and fine-tuned on a 2k sample of manually annotated Reddit (EN) and Twitter (PL) data from a political domain (reactions to political debate). Model developed as part of collaboration in the ComPathos project.
BERT-base sentiment classifier in Polish [PaReS-sentimenTw-political-PL]
The model allows to detect 3 categories of sentiment (negative, neutral, positive) in text. Developed for Polish language with the BERT-base-polish-cased model and fine-tuned on tweets from political domain (reactions to political debate). Model developed as part of collaboration in the ComPathos project.