Mining and Clustering Phrases for English for Special Purpose: Travel Writing

We introduce a method for learning to find paraphrases in a given corpus. In our approach, phrases are transformed into word embeddings aimed at finding the most close phrases to retrieve phrases based on machine learning. The method involves using language models to generate the word vector representation and find paraphrases. At run-time, phrases are transformed into a set of word embeddings, and re-ranking is performed on the retrieved paraphrases. We present a prototype model that applies the method to find common phrases and their paraphrases in a travel guide. Experimental evaluation on Lonely Planet Taiwan (Country Guide)[7] shows promising results.