This paper describes the DUTh's participation in the TREC 2020 Conversational Assistance Track (CAsT) track. Our approach incorporates linguistic analysis of the available queries along with query reformulation. The linguistic perspective of our approach implements the AllenNLP co-reference resolution model to every query of each conversational session. In addition, the SpaCy model was used for part-of-speech tagging and keyword extraction from the current and the previous turns. We reformulate the initial query into a weighted new query by keeping the keywords from the current turn and adding conversational context from previous turns. We argue that the conversational context of previous turns to have less impact than the keywords from the current turn while still adding some informational value. Finally, the new query was used for retrieval using Indri.