Inference Time Style Control for Summarization

Cao, Shijie; Wang, Lu

doi:10.18653/v1/2021.naacl-main.476

Cited by 10 publications

(5 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In doing so, we also introduce an effective method to create synthetic datasets for aspect-guided opinion summarization. Our work also relates to approaches which attempt to control summarization output based on length (Kikuchi et al, 2016), content (Fan et al, 2018), style (Cao and Wang, 2021), or textual queries (Dang, 2006). Although we focus solely on aspect, our method is general and could be used to adjust additional properties of a summary such as sentiment (e.g., positive vs. negative) or style (e.g., formal vs. colloquial).…”

Section: Related Workmentioning

confidence: 99%

Aspect-Controllable Opinion Summarization

Amplayo¹,

Angelidis²,

Lapata³

2021

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing

View full text Add to dashboard Cite

Recent work on opinion summarization produces general summaries based on a set of input reviews and the popularity of opinions expressed in them. In this paper, we propose an approach that allows the generation of customized summaries based on aspect queries (e.g., describing the location and room of a hotel). Using a review corpus, we create a synthetic training dataset of (review, summary) pairs enriched with aspect controllers which are induced by a multi-instance learning model that predicts the aspects of a document at different levels of granularity. We fine-tune a pretrained model using our synthetic dataset and generate aspect-specific summaries by modifying the aspect controllers. Experiments on two benchmarks show that our model outperforms the previous state of the art and generates personalized summaries by controlling the number of aspects discussed in them. HUMAN summariesGeneral Staff was service focused and very welcoming. Common areas of the hotel smelled fresh because of how clean everything was. The rooms were comfortable and came with a fridge and a microwave. Food, both hot and cold, was very well presented and fresh. The hotel was located within walking distance to the French quarter and felt very safe at night. Building It's older, looking at the hotel and lobby, but has lots of charm & character. Cleanliness The hotel's lounge, bathrooms, hallways, and even the bedding were all clean and even smelled fresh. Food The breakfast is very good and plentiful and was more than just continental, offering eggs, sausage and grits in addition to the usual waffles, cereal, and fruit. Location The location is very good, walking distance to all major sights in French quarter. Rooms The room is comfortable and equipped with just about everything anyone could need ... a refrigerator, microwave, desk, sofa, iron and ironing board, and hairdryer. The room was also spacious and the hotel was very quiet. Service Hotel staff were unbelievably friendly and helpful; they often went above and beyond to be accommodating. ACESUM summaries GeneralThe hotel is in a great location, close to the French quarter and the market. The room was clean and comfortable. Breakfast was good, and the staff was very helpful. There is a small restaurant in the lobby. Building The lobby is a bit small. The lobby area is a little bit dated, but the rooms are very comfortable. Cleanliness The room was clean and comfortable. The bathroom was very clean with a nice shower. Food The breakfast was very good, with a variety of choices. The breakfast buffet was good. Location The location is great, right in the heart of Bourbon street, and within walking distance of the French quarter. Rooms The room was very spacious and the bathroom was very nice. The room had a TV, a microwave, and a separate shower. There was a small fridge in the room, which was nice. Service The staff was very friendly and helpful.

show abstract

Section: Related Workmentioning

confidence: 99%

Aspect-Controllable Opinion Summarization

Amplayo¹,

Angelidis²,

Lapata³

2021

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing

View full text Add to dashboard Cite

show abstract

“…Diverse text generation has been studied in previous work (Yu et al, 2017), including in dialogue (Li et al, 2016), story generation (Fan et al, 2019), and particularly paraphrasing (Iyyer et al, 2018;Goyal and Durrett, 2020). Our method can also diversify content coverage (Gehrmann et al, 2018) and word choice (Cao and Wang, 2021).…”

Section: Related Workmentioning

confidence: 99%

Massive-scale Decoding for Text Generation using Lattices

Xu¹,

Jonnalagadda²,

Durrett³

2022

Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Langua

View full text Add to dashboard Cite

Conditional neural text generation models generate high-quality outputs, but often concentrate around a mode when what we really want is a diverse set of options. We present a search algorithm to construct lattices encoding a massive number of generation options. First, we restructure decoding as a best-first search, which explores the space differently than beam search and improves efficiency by avoiding pruning paths. Second, we revisit the idea of hypothesis recombination: we can identify pairs of similar generation candidates during search and merge them as an approximation. On both summarization and machine translation, we show that our algorithm encodes thousands of diverse options that remain grammatical and high-quality into one lattice. This algorithm provides a foundation for building downstream generation applications on top of massive-scale diverse outputs. 1

show abstract

“…Evaluating Alignment Algorithm. We search the hyperparameters on the Basil dataset (Fan et al, 2019) and test the algorithm on the Allsides dataset collected in Cao and Wang (2021). The Allsides dataset consists of manually aligned news articles from 251 media outlets.…”

Section: Appendix a Bignews Cleaning Stepsmentioning

confidence: 99%

POLITICS: Pretraining with Same-story Article Comparison for Ideology Prediction and Stance Detection

Liu¹,

Zhang²,

Wegsman³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Ideology is at the core of political science research. Yet, there still does not exist generalpurpose tools to characterize and predict ideology across different genres of text. To this end, we study Pretrained Language Models using novel ideology-driven pretraining objectives that rely on the comparison of articles on the same story written by media of different ideologies. We further collect a large-scale dataset, consisting of more than 3.6M political news articles, for pretraining. Our model POLITICS outperforms strong baselines and the previous state-of-the-art models on ideology prediction and stance detection tasks. Further analyses show that POLITICS is especially good at understanding long or formally written texts, and is also robust in few-shot learning scenarios.

show abstract

Inference Time Style Control for Summarization

Cited by 10 publications

References 13 publications

Aspect-Controllable Opinion Summarization

Aspect-Controllable Opinion Summarization

Massive-scale Decoding for Text Generation using Lattices

POLITICS: Pretraining with Same-story Article Comparison for Ideology Prediction and Stance Detection

Contact Info

Product

Resources

About