Quality of experience (QoE) cannot be evaluated only by simple quality measurement, since it is related to user's subjective evaluation, the specific application, purpose of use, etc. On the other hand, it is expected that the remarks on social network service (SNS) probably include the user's subjectivity. Estimating QoE from SNS remarks data has been considered in existing research. One problem in existing research is that communication quality related negative expression search is implemented by extracting negative words uniformly from a negative expression dictionary. Then, the meaning of each tweet is difficult to analyze. Therefore, user's subjectivity or QoE, cannot be estimated by extracting only negative words from the user's tweet. In this paper, as the first step of QoE estimation research, similarity of tweets is evaluated. The possibility of QoE estimation from tweets on communication quality has been verified through experiments with 168 tweets from 20 subjects. The experiment results show that it is possible to estimate QoE from tweets.