Document Summarization (3/3) Some studies combine extractive and abstractive method. 4 From: Get To The Point: Summarization with Pointer-Generator Networks (See et al. 2017) soft switch to choose extractive or abstractive
Summary ◼ A novel model for document summarization Model extracts sentences first, then rewrites them Use RL to bridge the nondifferentiable computation ◼ Improve repetition problem At sentence-level, this problem doesn’t occur because model extracts original sentences first At summary, it is improved by Repetition-Avoiding Reranking ◼ Get faster training and inference speed by parallel decoding ◼ SotA on CNN/Daily mail dataset 5
Summary ◼ A novel model for document summarization Model extracts sentences first, then rewrites them Use RL to bridge the nondifferentiable computation ◼ Improve repetition problem At sentence-level, this problem doesn’t occur because model extracts original sentences first At summary, it is improved by Repetition-Avoiding Reranking ◼ Get faster training and inference speed by parallel decoding ◼ SotA on CNN/Daily mail dataset 6
Model Architecture - Extractor 1. the convolutional encoder computes for each sentence 2. the RNN encoder (blue) computes ℎ 3. the RNN decoder (green) selects sentence at time step 9 ℎ4 ℎ3 ℎ2 ℎ1
Model Architecture - Abstractor Abstractor is given a sentence and rewrites it. The architecture is almost the same as below. 11 From: Get To The Point: Summarization with Pointer-Generator Networks (See et al. 2017)
Learning Starting from an end-to-end fashion is infeasible. When randomly initialized, the extractor would often select sentences that are not relevant, so it would be difficult for the abstractor to learn. Hence, learning is conducted in two phases: 1. Train the Extractor and Abstractor respectively 2. Reinforce-Guided Extraction (train the full model) 12
Learning – Extractor 1. Train the Extractor and Abstractor respectively Extractor ◼ Provide a ‘proxy’ target label by finding the most similar document sentence with ground-truth summary :
= argmax ROUGE– Lrecall
, ◼ Then trained to minimize the cross-entropy loss. 13 1 1 is most similar to … 3 2 is most similar to … 4 3 is most similar to … Ground-truth summary ‘proxy’ target labels
Learning – Abstractor 1. Train the Extractor and Abstractor respectively Abstractor ◼ Create training pairs by taking each summary sentence and pairing it with its extracted document sentence from previous method. ◼ The network is trained as an usual sequence-to-sequence model to minimize the cross-entropy loss . 14 1 1 3 2 4 3 Ground-truth summary ‘proxy’ target labels Abstractor
Learning – Full Model 2. Reinforce-Guided Extraction Extractor’s selecting sentence behavior is non-differentiable because the extraction probability of already extracted sentences are forced to zero. Use Reinforcement Learning by making the extractor an agent ◼ adopt Advantage Actor-Critic (A2C) State, Action, Reward are defined as: 16 State: = , −1 Action: ~ , ( , ) = () Reward: + 1 = ROUGE– LF1
, Document and extracted sentence at time step t-1 extraction probability from the Extractor Similarity between model output and ground-truth
Other Techniques ◼ Learning how many sentences to extract In the RL training phase, add another set of trainable parameters (EOE stands for ‘End-Of-Extraction’) ◼ Repetition-Avoiding Reranking At sentence-level, repetition problem doesn’t occur because model extracts original sentences first For removing a few ‘across-sentence’ repetitions, apply the same beam-search tri-gram avoidance 17
Experiment ◼ Human Evaluation Human answered “A is better, B is better, both are good/bad” ◼ Abstractiveness the ratio of novel n-grams in the generated summary 19
tatsui
kitachan_black
atamaplus
hiroki13
cutcarbon
inoichan
cocomoff
lancers_pr
ghtaro
shurain
kazaneya
ebisawahayata
yeseniaperezcruz
jasonvnalue
hatefulcrawdad
wjessup
addyosmani
zakiyama
tanoku
bkeepers
jrom
aki_iinuma
michaelherold
iamctodd