The MSR project focuses on developing novel techniques for text summarization. More specifically, we are interested in dialogue summarization, long-input summarization and few-shot summarization. Previously, we have created the QMSum, the first query-based dialogue summarization dataset as well as an exploratory study on QMSum and other dialogue summarization datasets. We are also developing methods for long-input summarization, using dynamic latent retrieval (DYLE), or multi-stage summarization framework (Summ^N). We are currently exploring the direction of using prompt learning for few-shot summarization. Apart from our research contributions, we have also developed SummerTime, an open-source text summarization toolkit for non-expert users, for which could help the users quickly explore different options for summarization datasets, models and evaluation metrics.