Answer selection(AS)is an important subtask of question answering(QA)that aims to choose the most suitable answer from a list of candidate an-swers.Existing AS models usually explored the single-scale sentence matching,whereas a sentence might contain semantic information at different scales,e.g.Word-level,Phrase-level,or the whole sentence.In addition,these models typi-cally use fixed-size feature vectors to represent questions and answers,which may cause information loss when questions or answers are too long.To address these issues,we propose an Encoder-Decoder Network with Cross-Match Mech-anism(EDCMN)where questions and answers that represented by feature vec-tors with fixed-size and dynamic-size are applied for multiple-perspective match-ing.In this model,Encoder layer is based on the “Siamese” network and Decoder layer is based on the “matching-aggregation” network.We evaluate our model on two tasks: Answer Selection and Textual Entailment.Experimental results show the effectiveness of our model,which achieves the state-of-the-art perfor-mance on WikiQA dataset.