In this paper, a theory named LI, which based on a fragment of linear logic, is put forward. In LI, effective results can be achieved when faced with inconsistent or non-monotonic information.