In April 2020, NIST proposed a first round of information retrieval evaluation including a set of topics of interest or queries and a collection of scientific documents related to COVID- 19. This paper presents our participation to this challenge by considering the first series of queries. Our model relies on the use of key phrases rather than the usual single words or roots. We used an n-gram based key phrase extraction model. We compared the performance of the search engine when documents are represented by single terms and by key phrases. We show that some queries benefit from the key term representation.