Paraphrase Identification Using Dependency Tree and Word Embeddings

In this paper, we are trying to develop an efficient and simple model for detecting paraphrase sentences in the English language. The dependency tree was chosen as the main structure to represent the relationships between words in a sentence. To represent the word semantics, we are using pre-trained general-purpose word embeddings. Based on these two key components, we designed a few features that can help to identify paraphrases. Conducted experiments proved that the model is efficient and shows relatively close results to state-of-the-art models.