Siamese Multiplicative LSTM for Semantic Text Similarity

Learning the Semantic Textual Similarity (STS) is a critical issue for many NLP tasks such as question answering, document summarization and etc.. In this paper, we combine the Multiplicative LSTM structure with a Siamese architecture which learn to project word embeddings of each sentence into a fixed-dimensional embedding space to represent this sentence. Then these sentence embeddings can be used to evaluate the STS task. We compare with several similar architectures and the proposed method has achieved better results and is competitive with the best state-of-the-art siamese neural network architecture.