Learning sentence embeddings using Recursive Networks
This work addresses the challenge of sentence representation for NLP applications, but it appears incremental as it compares existing methods without introducing a fundamentally new approach.
The paper tackled the problem of learning generalizable sentence embeddings by comparing three methods (LSTMs, recursive nets, and a POS-based variant) on dictionary definitions and the Rotten Tomatoes dataset, with results showing performance in a reverse dictionary application and transfer learning tasks.
Learning sentence vectors that generalise well is a challenging task. In this paper we compare three methods of learning phrase embeddings: 1) Using LSTMs, 2) using recursive nets, 3) A variant of the method 2 using the POS information of the phrase. We train our models on dictionary definitions of words to obtain a reverse dictionary application similar to Felix et al. [1]. To see if our embeddings can be transferred to a new task we also train and test on the rotten tomatoes dataset [2]. We train keeping the sentence embeddings fixed as well as with fine tuning.