n this presentation, I will start from the previous work I did in my master thesis and then some potential directions which I may work on during my PhD. The task of textual entailment can be roughly described as determining a directional relationship between two text fragments, Text(T) and Hypothesis(H), whether H can be infered from T. In the main approach of my master thesis, I have proposed a novel feature representation extracted from the dependency structure and then applied kernel-based machine learning techniques based on the entailment patterns. Two simple backup strategies have been considered to deal with the cases not covered by the main approach. After a detailed error analysis on the previous experiments and also others' work, I have found that the low coverage of the "deeper" methods is due to the mixture of several cases of entailment in the data sets, in which many linguistic phenomena are involved. Therefore, I reconsider the definition of entailment and also the classification of the data, in order to learn the essential characteristics of the problem. A possible framework is then proposed together with a discussion on the candidate linguistic representations. In addition, some potential applications based on textual entailment will be shown. At last, I will make a summary and propose a work plan.