EcForest: Extractive document summarization through enhanced sentence embedding and cascade forest

摘要

We present EcForest, an extractive summarization model through Enhanced SentenceEmbedding and Cascade Forest. Sentence representation is of great significance for many summarizationmethods. Bag-of-words mostly fails to grasp the semantics, and typical embeddingmodels cannot capture more complex semantic features, such as polysemy and the meaningof a phrase, which is usually ignored by simply averaging the word embeddings included in asentence. To this end, we propose Enhanced Sentence Embedding (ESE) model to solve suchdrawbacks via mapping several valid features to dense vectors. Essentially, the enhanced sentenceembedding is a novel model for improving the distributed representation of sentence.Our sentence embedding model is universally applicable and it can be adapted to other NLPtasks. Moreover, deep forest is used as a sentence extraction algorithm for its robustness tothe hyper-parameters and its efficient training algorithm compared to deep neural network.The evaluation of variant models proposed in this work proves the validation of the enhancedsentence embedding. The comparison results between EcForest and several baselines on twodifferent datasets demonstrate that the proposed summarization model performs better thanor with high competitiveness to the state-of-the-art.