Guided by grammatical structure, words compose to form sentences, and guided by discourse structure, sentences compose to form dialogues and documents. The compositional aspect of sentence and discourse units is often overlooked by machine learning algorithms. A recent initiative called Quantum Natural Language Processing (QNLP) learns word meanings as points in a Hilbert space and acts on them via a translation of grammatical structure into Parametrised Quantum Circuits (PQCs). Previous work extended the QNLP translation to discourse structure using points in a closure of Hilbert spaces. In this paper, we evaluate this translation on a Winograd-style pronoun resolution task. We train a Variational Quantum Classifier (VQC) for binary classification and implement an end-to-end pronoun resolution system. The simulations executed on IBMQ software converged with an F1 score of 87.20%. The model outperformed two out of three classical coreference resolution systems and neared state-of-the-art SpanBERT. A mixed quantum-classical model yet improved these results with an F1 score increase of around 6%.

Recent work on extending coreference resolution across domains and languages relies on annotated data in both the target domain and language. At the same time, pre-trained large language models (LMs) have been reported to exhibit strong zero- and few-shot learning abilities across a wide range of NLP tasks. However, prior work mostly studied this ability using artificial sentence-level datasets such as the Winograd Schema Challenge. In this paper, we assess the feasibility of prompt-based coreference resolution by evaluating instruction-tuned language models on difficult, linguistically-complex coreference benchmarks (e.g., CoNLL-2012). We show that prompting for coreference can outperform current unsupervised coreference systems, although this approach appears to be reliant on high-quality mention detectors. Further investigations reveal that instruction-tuned LMs generalize surprisingly well across domains, languages, and time periods; yet continued fine-tuning of neural models should still be preferred if small amounts of annotated examples are available.

Do question answering (QA) modeling improvements (e.g., choice of architecture and training procedure) hold consistently across the diverse landscape of QA benchmarks? To study this question, we introduce the notion of concurrence -- two benchmarks have high concurrence on a set of modeling approaches if they rank the modeling approaches similarly. We measure the concurrence between 32 QA benchmarks on a set of 20 diverse modeling approaches and find that human-constructed benchmarks have high concurrence amongst themselves, even if their passage and question distributions are very different. Surprisingly, even downsampled human-constructed benchmarks (i.e., collecting less data) and programmatically-generated benchmarks (e.g., cloze-formatted examples) have high concurrence with human-constructed benchmarks. These results indicate that, despite years of intense community focus on a small number of benchmarks, the modeling improvements studied hold broadly.

Transformer models cannot easily scale to long sequences due to their O(N^2) time and space complexity. This has led to Transformer variants seeking to lower computational complexity, such as Longformer and Performer. While such models have theoretically greater efficiency, their effectiveness on real NLP tasks has not been well studied. We benchmark 7 variants of Transformer models on 5 difficult NLP tasks and 7 datasets. We design experiments to isolate the effect of pretraining and hyperparameter settings, to focus on their capacity for long-range attention. Moreover, we present various methods to investigate attention behaviors to illuminate model details beyond metric scores. We find that the modified attention in long-range transformers has advantages on content selection and query-guided decoding, but they come with previously unrecognized drawbacks such as insufficient attention to distant tokens and accumulated approximation error.

Predicting emotions expressed in text is a well-studied problem in the NLP community. Recently there has been active research in extracting the cause of an emotion expressed in text. Most of the previous work has done causal emotion entailment in documents. In this work, we propose neural models to extract emotion cause span and entailment in conversations. For learning such models, we use RECCON dataset, which is annotated with cause spans at the utterance level. In particular, we propose MuTEC, an end-to-end Multi-Task learning framework for extracting emotions, emotion cause, and entailment in conversations. This is in contrast to existing baseline models that use ground truth emotions to extract the cause. MuTEC performs better than the baselines for most of the data folds provided in the dataset.

To promote and further develop RST-style discourse parsing models, we need a strong baseline that can be regarded as a reference for reporting reliable experimental results. This paper explores a strong baseline by integrating existing simple parsing strategies, top-down and bottom-up, with various transformer-based pre-trained language models. The experimental results obtained from two benchmark datasets demonstrate that the parsing performance strongly relies on the pretrained language models rather than the parsing strategies. In particular, the bottom-up parser achieves large performance gains compared to the current best parser when employing DeBERTa. We further reveal that language models with a span-masking scheme especially boost the parsing performance through our analysis within intra- and multi-sentential parsing, and nuclearity prediction.

Centering theory (CT; Grosz et al., 1995) provides a linguistic analysis of the structure of discourse. According to the theory, local coherence of discourse arises from the manner and extent to which successive utterances make reference to the same entities. In this paper, we investigate the connection between centering theory and modern coreference resolution systems. We provide an operationalization of centering and systematically investigate if neural coreference resolvers adhere to the rules of centering theory by defining various discourse metrics and developing a search-based methodology. Our information-theoretic analysis reveals a positive dependence between coreference and centering; but also shows that high-quality neural coreference resolvers may not benefit much from explicitly modeling centering ideas. Our analysis further shows that contextualized embeddings contain much of the coherence information, which helps explain why CT can only provide little gains to modern neural coreference resolvers which make use of pretrained representations. Finally, we discuss factors that contribute to coreference which are not modeled by CT such as world knowledge and recency bias. We formulate a version of CT that also models recency and show that it captures coreference information better compared to vanilla CT.

Sluice resolution is a problem where a system needs to output the corresponding antecedents of wh-ellipses. The antecedents are elided contents behind the wh-words but are implicitly referred to using contexts. Previous work frames sluice resolution as question answering where this setting outperforms all its preceding works by large margins. Ellipsis and questions are referentially dependent expressions (anaphoras) and retrieving the corresponding antecedents are like answering questions to output pieces of clarifying information. However, the task is not fully solved. Therefore, we want to further investigate what makes sluice resolution differ to question answering and fill in the error gaps. We also present some results using recent state-of-the-art question answering systems which improve the previous work (86.01 to 90.39 F1).