COLING 2024 Papers

Privacy policy documents have a crucial role in educating individuals about the collection, usage, and protection of users’ personal data by organizations. However, they are notorious for their lengthy, complex, and convoluted language especially involving privacy-related entities. Hence, they pose a significant challenge to users who attempt to comprehend organization’s data usage policy. In this paper, we propose to enhance the interpretability and readability of policy documents by using controlled abstractive summarization – we enforce the generated summaries to include critical privacy-related entities (e.g., data and medium) and organization’s rationale (e.g., target and reason) in collecting those entities. To achieve this, we develop PD-Sum, a policy-document summarization dataset with marked privacy-related entity labels. Our proposed model, EROS, identifies critical entities through a span-based entity extraction model and employs them to control the information content of the summaries using proximal policy optimization (PPO). Comparison shows encouraging improvement over various baselines. Furthermore, we furnish qualitative and human evaluations to establish the efficacy of EROS.

Translation quality estimation (QE) is an important component in real-world machine translation applications. Unfortunately, human labeled QE datasets, which play an important role in developing and assessing QE models, are only available for limited language pairs. In this paper, we present the first English-Persian QE dataset, called EPOQUE, which has manually annotated direct assessment labels. EPOQUE contains 1000 sentences translated from English to Persian and annotated by three human annotators. It is publicly available, and thus can be used as a zero-shot test set, or for other scenarios in future work. We also evaluate and report the performance of two state-of-the-art QE models, i.e., Transquest and CometKiwi, as baselines on our dataset. Furthermore, our experiments show that using a small subset of the proposed dataset containing 300 sentences to fine-tune Transquest, can improve its performance by more that 8% in terms of the Pearson correlation with a held-out test set.

We propose EMOLIS Dataset that contains annotated emotional transcripts of scenes from Walt Disney cartoons at the same time as physiological signals from spectators (breathing, ECG, eye movements). The dataset is used in EMOLIS App, our second proposal. EMOLIS App allows to display the identified emotions while a video is playing and suggest emotionally comparable videos. We propose to estimate an emotional distance between videos using multimodal neural representations (text, audio, video) that also combine physiological signals. This enables personalized results that can be used for cognitive therapies focusing on awareness of felt emotions. The dataset is designed to be suitable for all audiences and autistic people who have difficulties to recognize and express emotions.

There have been many attempts to model the morphological richness and complexity of Arabic, leading to numerous Part-of-Speech (POS) tagsets that differ in terms of (a) which morphological features they represent, (b) how they represent them, and (c) the degree of specification of said features. Tagset granularity plays an important role in determining how annotated data can be used and for what applications. Due to the diversity among existing tagsets, many annotated corpora for Arabic cannot be easily combined, which exacerbates the Arabic resource poverty situation. In this work, we propose an intermediate tagset designed to facilitate the conversion and unification of different tagsets used to annotate Arabic corpora. This new tagset acts as a bridge between different annotation schemes, simplifying the integration of annotated corpora and promoting collaboration across the projects using them.

In this work, we revisit the problem of semi-supervised named entity recognition (NER) focusing on extremely light supervision, consisting of a lexicon containing only 10 examples per class. We introduce ELLEN, a simple, fully modular, neuro-symbolic method that blends fine-tuned language models with linguistic rules. These rules include insights such as “One Sense Per Discourse”, using a Masked Language Model as an unsupervised NER, leveraging part-of-speech tags to identify and eliminate unlabeled entities as false negatives, and other intuitions about classifier confidence scores in local and global context. ELLEN achieves very strong performance on the CoNLL-2003 dataset when using the minimal supervision from the lexicon above. It also outperforms most existing (and considerably more complex) semi-supervised NER methods under the same supervision settings commonly used in the literature (i.e., 5% of the training data). Further, we evaluate our CoNLL-2003 model in a zero-shot scenario on WNUT-17 where we find that it outperforms GPT-3.5 and achieves comparable performance to GPT-4. In a zero-shot setting, ELLEN also achieves over 75% of the performance of a strong, fully supervised model trained on gold data. Our code is publicly available.

Transformer-based models have demonstrated outstanding performance in natural language processing (NLP) tasks and many other domains, e.g., computer vision. Depending on the size of these models, which have grown exponentially in the past few years, machine learning practitioners might be restricted from deploying them in resource-constrained environments. This paper discusses the compression of transformer-based models for multiple resource budgets. Integrating neural architecture search (NAS) and network pruning techniques, we effectively generate and train weight-sharing super-networks that contain efficient, high-performing, and compressed transformer-based models. A common challenge in NAS is the design of the search space, for which we propose a method to automatically obtain the boundaries of the search space and then derive the rest of the intermediate possible architectures using a first-order weight importance technique. The proposed end-to-end NAS solution, EFTNAS, discovers efficient subnetworks that have been compressed and fine-tuned for downstream NLP tasks. We demonstrate EFTNAS on the General Language Understanding Evaluation (GLUE) benchmark and the Stanford Question Answering Dataset (SQuAD), obtaining high-performing smaller models with a reduction of more than 5x in size without or with little degradation in performance.

Current approaches to question answering rely on pre-trained language models (PLMs) like RoBERTa. This work challenges the existing question-answer encoding convention and explores finer representations. We begin with testing various pooling methods compared to using the begin-of-sentence token as a question representation for better quality. Next, we explore opportunities to simultaneously embed all answer candidates with the question. This enables cross-reference between answer choices and improves inference throughput via reduced memory usage. Despite their simplicity and effectiveness, these methods have yet to be widely studied in current frameworks. We experiment with different PLMs, and with and without the integration of knowledge graphs. Results prove that the memory efficacy of the proposed techniques with little sacrifice in performance. Practically, our work enhances 38-100% throughput with 26-65% speedups on consumer-grade GPUs by allowing for considerably larger batch sizes. Our work sends a message to the community with promising directions in both representation quality and efficiency for the question-answering task in natural language processing.

We present EDEN, the first Norwegian dataset annotated with event information at the sentence level, adapting the widely used ACE event schema to Norwegian. The paper describes the manual annotation of Norwegian text as well as transcribed speech in the news domain, together with inter-annotator agreement and discussions of relevant dataset statistics. We also present preliminary modeling results using a graph-based event parser. The resulting dataset will be freely available for download and use.

Stance detection aims to determine the attitude expressed in text towards a given target. Zero-shot stance detection (ZSSD) has emerged to classify stances towards unseen targets during inference. Recent data augmentation techniques for ZSSD increase transferable knowledge between targets through text or target augmentation. However, these methods exhibit limitations. Target augmentation lacks logical connections between generated targets and source text, while text augmentation relies solely on training data, resulting in insufficient generalization. To address these issues, we propose an encoder-decoder data augmentation (EDDA) framework. The encoder leverages large language models and chain-of-thought prompting to summarize texts into target-specific if-then rationales, establishing logical relationships. The decoder generates new samples based on these expressions using a semantic correlation word replacement strategy to increase syntactic diversity. We also analyze the generated expressions to develop a rationale-enhanced network that fully utilizes the augmented data. Experiments on benchmark datasets demonstrate our approach substantially improves over state-of-the-art ZSSD techniques. The proposed EDDA framework increases semantic relevance and syntactic variety in augmented texts while enabling interpretable rationale-based learning.

In common law jurisdictions, legal practitioners rely on precedents to construct arguments, in line with the doctrine of stare decisis. As the number of cases grow over the years, prior case retrieval (PCR) has garnered significant attention. Besides lacking real-world scale, existing PCR datasets do not simulate a realistic setting, because their queries use complete case documents while only masking references to prior cases. The query is thereby exposed to legal reasoning not yet available when constructing an argument for an undecided case as well as spurious patterns left behind by citation masks, potentially short-circuiting a comprehensive understanding of case facts and legal principles. To address these limitations, we introduce a PCR dataset based on judgements from the European Court of Human Rights (ECtHR), which explicitly separate facts from arguments and exhibit precedential practices, aiding us to develop this PCR dataset to foster systems’ comprehensive understanding. We benchmark different lexical and dense retrieval approaches with various negative sampling strategies, adapting them to deal with long text sequences using hierarchical variants. We found that difficulty-based negative sampling strategies were not effective for the PCR task, highlighting the need for investigation into domain-specific difficulty criteria. Furthermore, we observe performance of the dense models degrade with time and calls for further research into temporal adaptation of retrieval models. Additionally, we assess the influence of different views , Halsbury’s and Goodhart’s, in practice in ECtHR jurisdiction using PCR task.

Skeleton-aware sign language recognition (SLR) has gained popularity due to its ability to remain unaffected by background information and its lower computational requirements. Current methods utilize spatial graph modules and temporal modules to capture spatial and temporal features, respectively. However, their spatial graph modules are typically built on fixed graph structures such as graph convolutional networks or a single learnable graph, which only partially explore joint relationships. Additionally, a simple temporal convolution kernel is used to capture temporal information, which may not fully capture the complex movement patterns of different signers. To overcome these limitations, we propose a new spatial architecture consisting of two concurrent branches, which build input-sensitive joint relationships and incorporates specific domain knowledge for recognition, respectively. These two branches are followed by an aggregation process to distinguishe important joint connections. We then propose a new temporal module to model multi-scale temporal information to capture complex human dynamics. Our method achieves state-of-the-art accuracy compared to previous skeleton-aware methods on four large-scale SLR benchmarks. Moreover, our method demonstrates superior accuracy compared to RGB-based methods in most cases while requiring much fewer computational resources, bringing better accuracy-computation trade-off. Code is available at https://github.com/hulianyuyy/DSTA-SLR.

In this paper, we study the problem of multi-reward reinforcement learning to jointly optimize for multiple text qualities for natural language generation. We focus on the task of counselor reflection generation, where we optimize the generators to simultaneously improve the fluency, coherence, and reflection quality of generated counselor responses. We introduce two novel bandit methods, DynaOpt and C-DynaOpt, which rely on the broad strategy of combining rewards into a single value and optimizing them simultaneously. Specifically, we employ non-contextual and contextual multi-arm bandits to dynamically adjust multiple reward weights during training. Through automatic and manual evaluations, we show that our proposed techniques, DynaOpt and C-DynaOpt, outperform existing naive and bandit baselines, showcasing their potential for enhancing language models.

Automatic generation of radiology reports can relieve the burden of radiologist. In the radiology library, the biased dataset and the sparse features of chest X-ray image make it difficult to generate reports. Many approaches strive to integrate prior information to enhance generation, but they fail to dynamically utilize pulmonary lesion knowledge at the instance-level. To alleviate above problem, we propose a novel Dynamic Knowledge Prompt (DKP) framework for chest X-ray report generation. The DKP can dynamically incorporate the pulmonary lesion information at the instance-level to facilitate report generation. Initially, we design a knowledge prompt for each pulmonary lesion using numerous radiology reports. After that, the DKP using an anomaly detector generates the dynamic knowledge prompt by extracting discriminative lesion features in the corresponding X-ray image. Finally, the knowledge prompt is encoded and fused with hidden states extracted from decoder, to form multi-modal features that guide visual features to generate reports. Extensive experiments on the public datasets MIMIC-CXR and IU X-Ray show that our approach achieves state-of-the-art performance.

User Simulators play a pivotal role in training and evaluating task-oriented dialogue systems. Traditional user simulators typically rely on human-engineered agendas, resulting in generated responses that often lack diversity and spontaneity. Although large language models (LLMs) exhibit a remarkable capacity for generating coherent and contextually appropriate utterances, they may fall short when tasked with generating responses that effectively guide users towards their goals, particularly in dialogues with intricate constraints and requirements. This paper introduces DuetSim, a novel framework designed to address the intricate demands of task-oriented dialogues by leveraging LLMs. DuetSim stands apart from conventional approaches by employing two LLMs in tandem: one dedicated to response generation and the other focused on verification. This dual LLM approach empowers DuetSim to produce responses that not only exhibit diversity but also demonstrate accuracy and are preferred by human users. We validate the efficacy of our method through extensive experiments conducted on the MultiWOZ dataset, highlighting improvements in response quality and correctness, largely attributed to the incorporation of the second LLM.

Aspect Sentiment Triple Extraction (ASTE) is an emerging task in fine-grained sentiment analysis. Recent studies have employed Graph Neural Networks (GNN) to model the syntax-semantic relationships inherent in triplet elements. However, they have yet to fully tap into the vast potential of syntactic and semantic information within the ASTE task. In this work, we propose a Dual Encoder: Exploiting the potential of Syntactic and Semantic model (D2E2S), which maximizes the syntactic and semantic relationships among words. Specifically, our model utilizes a dual-channel encoder with a BERT channel to capture semantic information, and an enhanced LSTM channel for comprehensive syntactic information capture. Subsequently, we introduce the heterogeneous feature interaction module to capture intricate interactions between dependency syntax and attention semantics, and to dynamically select vital nodes. We leverage the synergy of these modules to harness the significant potential of syntactic and semantic information in ASTE tasks. Testing on public benchmarks, our D2E2S model surpasses the current state-of-the-art(SOTA), demonstrating its effectiveness.

Knowledge graph embedding, which aims to learn representations of entities and relations in large scale knowledge graphs, plays a crucial part in various downstream applications. The performance of knowledge graph embedding models mainly depends on the ability of modeling relation patterns, such as symmetry/antisymmetry, inversion and composition (commutative composition and non-commutative composition). Most existing methods fail in modeling the non-commutative composition patterns. Several methods support this kind of pattern by modeling in quaternion space or dihedral group. However, extending to such sophisticated spaces leads to a substantial increase in the amount of parameters, which greatly reduces the parameter efficiency. In this paper, we propose a new knowledge graph embedding method called dual complex number knowledge graph embeddings (DCNE), which maps entities to the dual complex number space, and represents relations as rotations in 2D space via dual complex number multiplication. The non-commutativity of the dual complex number multiplication empowers DCNE to model the non-commutative composition patterns. In the meantime, modeling relations as rotations in 2D space can effectively improve the parameter efficiency. Extensive experiments on multiple benchmark knowledge graphs empirically show that DCNE achieves significant performance in link prediction and path query answering.

The biomedical domain has sparked a significant interest in the field of Natural Language Processing (NLP), which has seen substantial advancements with pre-trained language models (PLMs). However, comparing these models has proven challenging due to variations in evaluation protocols across different models. A fair solution is to aggregate diverse downstream tasks into a benchmark, allowing for the assessment of intrinsic PLMs qualities from various perspectives. Although still limited to few languages, this initiative has been undertaken in the biomedical field, notably English and Chinese. This limitation hampers the evaluation of the latest French biomedical models, as they are either assessed on a minimal number of tasks with non-standardized protocols or evaluated using general downstream tasks. To bridge this research gap and account for the unique sensitivities of French, we present the first-ever publicly available French biomedical language understanding benchmark called DrBenchmark. It encompasses 20 diversified tasks, including named-entity recognition, part-of-speech tagging, question-answering, semantic textual similarity, or classification. We evaluate 8 state-of-the-art pre-trained masked language models (MLMs) on general and biomedical-specific data, as well as English specific MLMs to assess their cross-lingual capabilities. Our experiments reveal that no single model excels across all tasks, while generalist models are sometimes still competitive.

Open Domain Multi-Hop Question Answering (ODMHQA) plays a crucial role in Natural Language Processing (NLP) by aiming to answer complex questions through multi-step reasoning over retrieved information from external knowledge sources. Recently, Large Language Models (LLMs) have demonstrated remarkable performance in solving ODMHQA owing to their capabilities including planning, reasoning, and utilizing tools. However, LLMs may generate off-topic answers when attempting to solve ODMHQA, namely the generated answers are irrelevant to the original questions. This issue of off-topic answers accounts for approximately one-third of incorrect answers, yet remains underexplored despite its significance. To alleviate this issue, we propose the Discriminate→Re-Compose→Re- Solve→Re-Decompose (Dr3) mechanism. Specifically, the Discriminator leverages the intrinsic capabilities of LLMs to judge whether the generated answers are off-topic. In cases where an off-topic answer is detected, the Corrector performs step-wise revisions along the reversed reasoning chain (Re-Compose→Re-Solve→Re-Decompose) until the final answer becomes on-topic. Experimental results on the HotpotQA and 2WikiMultiHopQA datasets demonstrate that our Dr3 mechanism considerably reduces the occurrence of off-topic answers in ODMHQA by nearly 13%, improving the performance in Exact Match (EM) by nearly 3% compared to the baseline method without the Dr3 mechanism.

Few-shot relation extraction (FSRE) can alleviate the data scarcity problem in relation extraction. However, FSRE models often suffer a significant decline in performance when adapting to new domains. To overcome this issue, many researchers have focused on domain adaption FSRE (DAFSRE). Nevertheless, existing approaches primarily concentrate on the source domain, which makes it difficult to accurately transfer useful knowledge to the target domain. Additionally, the lack of distinction between relations further restricts the model performance. In this paper, we propose the domain-aware and co-adaptive feature transformation approach to address these issues. Specifically, we introduce a domain-aware transformation module that leverages the target domain distribution features to guide the domain-aware feature transformations. This can enhance the model’s adaptability across domains, leading to improved target domain performance. Furthermore, we design co-adaptive prototypical networks to perform co-adaptive feature transformation through a transformer mechanism. This results in more robust and distinguishable relation prototypes. Experiments on DAFSRE benchmark datasets demonstrate the effectiveness of our method, which outperforms existing models and achieves state-of-the-art performance.