CorefQA: Coreference Resolution as Query-based Span Prediction

By Wei Wu, Fei Wang, Arianna Yuan, Fei Wu, Jiwei Li. 2019

196 citations recorded by the source corpus.

In this paper, we present CorefQA, an accurate and extensible approach for the coreference resolution task. We formulate the problem as a span prediction task, like in question answering: A query is generated for each candidate mention using its surrounding context, and a span prediction module is employed to extract the text spans of the coreferences within the document using the generated query. This formulation comes with the following key advantages: (1) The span prediction strategy provides the flexibility of retrieving mentions left out at the mention proposal stage; (2) In the question answering framework, encoding the mention and its context explicitly in a query makes it possible to have a deep and thorough examination of cues embedded in the context of coreferent mentions; and (3) A plethora of existing question answering datasets can be used for data augmentation to improve the model’s generalization capability. Experiments demonstrate significant performance boost over previous models, with 83.1 (+3.5) F1 score on the CoNLL-2012 benchmark and 87.5 (+2.5) F1 score on the GAP benchmark.

Subfields: Artificial Intelligence, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Natural Language Processing

Citations in the corpus

Finetuned Language Models Are Zero-Shot Learners (2021, methodology): “Other work has used QA-based task formulation for more-targeted applications including semantic role labeling (He et al., 2015), relation extraction (Levy et al., 2017), coreference resolution (Wu et al., 2020) and named entity recognition (Li et al., 2020) as question answering.”
Unifying Large Language Models and Knowledge Graphs: A Roadmap (2023, methodology): “CorefQA [210] formulates coreference resolution as a question answering task, where contextual queries are generated for each candidate mention and the coreferent spans are extracted from the document using the queries.”
Example-Based Named Entity Recognition (2020, background): “For instance, (Wu et al., 2019) and (Li et al., 2019) have done this for coreference resolution and NER, respectively.”
Structured Prediction as Translation between Augmented Natural Languages (2021, methodology): “In particular, when the dataset sizes are small such as in ADE or CoNLL04, we obtain sizable improvements and (Luan et al., 2019) 88.4 63.2 MRC4ERE (Zhao et al., 2020) 88.9 71.9 85.5 62.1 RSAN (Yuan et al., 2020) 84 All results presented in this paper are obtained from a pre-trained T5-base model.”
Pre-training of Mention Representations in Coreference Models (2020, methodology): “However, the models are not completely comparable because Wu et al. (2019) pre-train on Quoref and SQUAD, whereas we pre-train on Wikipedia.”
DeepStruct: Pretraining of Language Models for Structure Prediction (2022, methodology): “…(Joshi et al., 2019): this model replaces original recurrent neural network backbone in (Lee et al., 2018) with BERT, (iii) CorefQA+SpanBERT (Wu et al., 2020): the model formulates coreference resolution as question answering, (iv) TANL (Paolini et al., 2021): this is a sequence-to-sequence…”
Description Based Text Classification with Reinforcement Learning (2020, background): “Wang et al. (2019) proposed to frame ABSA towards question answering (QA), and designed a reinforced attention network to select aspect-speciﬁc words, which alleviates the effects of noisy words for a speciﬁc aspect.”
How Optimal is Greedy Decoding for Extractive Question Answering? (2021, background): “, 2017], coreference resolution [Wu et al., 2020], named entity recognition [Li et al.”
Paired Representation Learning for Event and Entity Coreference (2020, methodology): “We compare with two baselines that stand as the state-of-the-art: • (Wu et al., 2020): formulating the task as query-based span prediction in question answering; • (Joshi et al., 2020): using a new pre-training representation SpanBERT as the embedding of the end-to-end coreference model by Lee et…”
Event Extraction as Machine Reading Comprehension (2020, background): “Previous works connecting MRC and other tasks usually adopt human-designed templates (Levy et al., 2017; FitzGerald et al., 2018; Li et al., 2019b,a; Gao et al., 2019; Wu et al., 2019).”
Recent Advances in Natural Language Processing via Large Pre-trained Language Models: A Survey (2021, methodology): “[200] Use question-answering to ind a coreferent mention of a marked mention from within the same text.”
Natural Language Processing Advancements By Deep Learning: A Survey (2020, background): “[138] reformulates coreference resolution as a span prediction task as in question answering and provide superior performance on the CoNLL2012 benchmark task.”