Service that processes threads of short texts such as tweets or forum postings and attempts to determine the stance (support, deny, question or comment) of each reply towards the post it is replying to. Input is a CSV with at least three columns; the text of each post, and identifier for that post, and in the case of replies the identifier of the post that is being replied to. Output is a CSV with the stance classification for each reply. This method uses a multilingual BERT-based model and is able to classify threads where the reply and target posts are in the same or different languages