Here is our submission for the third Touché lab at CLEF 2022, shared task 2 on Argument Retrieval for Comparative Questions, which tackles answering comparative questions based on argument retrieval of text passages to support answering comparative questions in the scenario of personal decision making. The previous two Touché editions mostly focused on retrieving complete arguments and documents, while this edition is about whether argument retrieval can support decision making directly by extracting the argumentative gist from documents, by classifying their stance with respect to the objects compared. The pipeline of our approach includes query pre-processing, query expansion, document retrieval using elastic search, and further re-rank them through argument identification which is based on SVM and DistilBERT-based stacked model and sorting the documents by multiple ranking scores and finally retrieving top relevant arguments for stance detection by embedded LSTM model.