COLING 2024 论文列表

For a viewpoint-diverse news recommender, identifying whether two news articles express the same viewpoint is essential. One way to determine “same or different” viewpoint is stance detection. In this paper, we investigate the robustness of operationalization choices for few-shot stance detection, with special attention to modelling stance across different topics. Our experiments test pre-registered hypotheses on stance detection. Specifically, we compare two stance task definitions (Pro/Con versus Same Side Stance), two LLM architectures (bi-encoding versus cross-encoding), and adding Natural Language Inference knowledge, with pre-trained RoBERTa models trained with shots of 100 examples from 7 different stance detection datasets. Some of our hypotheses and claims from earlier work can be confirmed, while others give more inconsistent results. The effect of the Same Side Stance definition on performance differs per dataset and is influenced by other modelling choices. We found no relationship between the number of training topics in the training shots and performance. In general, cross-encoding out-performs bi-encoding, and adding NLI training to our models gives considerable improvement, but these results are not consistent across all datasets. Our results indicate that it is essential to include multiple datasets and systematic modelling experiments when aiming to find robust modelling choices for the concept ‘stance’.

We outline the ongoing development of the Indiana Parsed Corpus of (Historical) High German. Once completed, this corpus will fill the gap in Penn-style treebanks for Germanic languages by spanning High German from 1050 to 1950. This paper describes the process of building the corpus: selection of texts, decisions on part-of-speech tags and other labels, the process of annotation, and illustrative annotation issues unique to historical High German. The construction of the corpus has led to a refinement of the Penn labels, tailored to the particulars of this language.

We present a new question answering corpus in French designed to educational domain. To be useful in such domain, we have to propose more complex questions and to be able to justify the answers on validated material. We analyze some properties of this corpus. The last part of this paper will be devoted to present the first experiments we have carried out to demonstrate the value of this dataset for learning a Retrieval Augmented Genration framework. Different experiments are proposed, with an automatic evaluation. A human evaluation is proposed to confirm or infirm this automatic evaluation.

The large success of deep learning based methods in Visual Question Answering (VQA) has concurrently increased the demand for explainable methods. Most methods in Explainable Artificial Intelligence (XAI) focus on generating post-hoc explanations rather than taking an intrinsic approach, the latter characterizing an interpretable model. In this work, we introduce an interpretable approach for graph-based VQA and demonstrate competitive performance on the GQA dataset. This approach bridges the gap between interpretability and performance. Our model is designed to intrinsically produce a subgraph during the question-answering process as its explanation, providing insight into the decision making. To evaluate the quality of these generated subgraphs, we compare them against established post-hoc explainability methods for graph neural networks, and perform a human evaluation. Moreover, we present quantitative metrics that correlate with the evaluations of human assessors, acting as automatic metrics for the generated explanatory subgraphs. Our code will be made publicly available at link removed due to anonymity period.

Reading comprehension continues to be a crucial research focus in the NLP community. Recent advances in Machine Reading Comprehension (MRC) have mostly centered on literal comprehension, referring to the surface-level understanding of content. In this work, we focus on the next level - interpretive comprehension, with a particular emphasis on inferring the themes of a narrative text. We introduce the first dataset specifically designed for interpretive comprehension of educational narratives, providing corresponding well-edited theme texts. The dataset spans a variety of genres and cultural origins and includes human-annotated theme keywords with varying levels of granularity. We further formulate NLP tasks under different abstractions of interpretive comprehension toward the main idea of a story. After conducting extensive experiments with state-of-the-art methods, we found the task to be both challenging and significant for NLP research. The dataset and source code have been made publicly available to the research community at https://github.com/RiTUAL-UH/EduStory.

With the rapid development of social media and short video applications in recent years, browsing short videos has become the norm. Due to its large user base and unique appeal, spreading rumors via short videos has become a severe social problem. Many methods simply fuse multimodal features for rumor detection, which lack interpretability. For short video rumors, rumor makers create rumors by modifying and/or splicing different modal information, so we should consider how to detect rumors from the perspective of modality tampering. Inspired by cross-modal contrastive learning, we propose a novel short video rumor detection framework by designing two pretraining tasks: modality tampering detection and inter-modal matching, imbuing the model with the ability to detect modality tampering and employing it for downstream rumor detection tasks. In addition, we design an interpretability mechanism to make the rumor detection results more reasonable by backtracking the model’s decision-making process. The experimental results show that the method on the short video rumor dataset has an improvement of about 4.6%-12% in macro-F1 compared with other models and can explain whether the short video is a rumor or not through the perspective of modality tampering.

This paper sheds light on a relatively unexplored area which is deep learning interpretability for speech disorder assessment and characterization. Building upon a state-of-the-art methodology for the explainability and interpretability of hidden representation inside a deep-learning speech model, we provide a deeper understanding and interpretation of the final intelligibility assessment of patients experiencing speech disorders due to Head and Neck Cancers (HNC). Promising results have been obtained regarding the prediction of speech intelligibility and severity of HNC patients while giving relevant interpretations of the final assessment both at the phonemes and phonetic feature levels. The potential of this approach becomes evident as clinicians can acquire more valuable insights for speech therapy. Indeed, this can help identify the specific linguistic units that affect intelligibility from an acoustic point of view and enable the development of tailored rehabilitation protocols to improve the patient’s ability to communicate effectively, and thus, the patient’s quality of life.

The notion of face described by Brown and Levinson (1987) has been studied in great detail, but a critical aspect of the framework, that which focuses on how intentions mediate the planning of turns which impose upon face, has received far less attention. We present an analysis of three computational systems trained for classifying both intention and politeness, focusing on how the former influences the latter. In politeness theory, agents attend to the desire to have their wants appreciated (positive face), and a complementary desire to act unimpeded and maintain freedom (negative face). Similar to speech acts, utterances can perform so-called face acts which can either raise or threaten the positive or negative face of the speaker or hearer. We begin by using an existing corpus to train a model which classifies face acts, achieving a new SoTA in the process. We then observe that every face act has an underlying intention that motivates it and perform additional experiments integrating dialog act annotations to provide these intentions by proxy. Our analysis finds that dialog acts improve performance on face act detection for minority classes and points to a close relationship between aspects of face and intent.

Counterspeech is an effective way to combat online hate speech. Considering the multifaceted nature of online hate speech, counterspeech with varying intents (e.g., denouncing or empathy) has significant potential to mitigate hate speech effectively. Recently, controlled approaches based on large language models (LLMs) have been explored to generate intent-specific counterspeech. Due to the lack of attention to intent-specific information by LLMs during the decoding process, those methods cater more to the semantic information rather than matching with the desired intents. Further, there are still limitations in quantitatively evaluating the effectiveness of counterspeech with different intents in mitigating hate speech. In this paper, to address the above issues, we propose DART, an LLMs-based DuAl-discRiminaTor guided framework for counterspeech generation. We employ an intent-aware discriminator and hate-mitigating discriminator to jointly guide the decoding preferences of LLMs, which facilitates the model towards generating counterspeech catering to specific intent and hate mitigation. We apply a maximum-margin relative objective for training discriminators. This objective leverages the distance between counterspeech aligned with the desired target (such as specific intent or effectiveness in hate mitigation) and undesired as an effective learning signal. Extensive experiments show that DART achieves excellent performances in matching the desired intent and mitigating hate.

Multimodal emotion recognition (MER) aims to identify emotions by utilizing affective information from multiple modalities. Due to the inherent disparities among these heterogeneous modalities, there is a large modality gap in their representations, leading to the challenge of fusing multiple modalities for MER. To address this issue, this work proposes a novel attention-based MER framework by integrating representation subspace mapping with unimodal auxiliary loss for enhancing multimodal fusion capabilities. Initially, a representation subspace mapping module is proposed to map each modality into two distinct subspaces. One is modality-public, enabling the acquisition of common representations and reducing the discrepancies across modalities. The other is modality-unique, retaining the unique characteristics of each modality while eliminating redundant inter-modal attributes. Then, a cross-modality attention is leveraged to bridge the modality gap in unique representations and facilitate modality adaptation. Additionally, our method designs an unimodal auxiliary loss to remove the noise unrelated to emotion classification, resulting in robust and meaningful representations for MER. Comprehensive experiments are conducted on the IEMOCAP and MSP-Improv datasets, and experiment results show that our method achieves superior performance to state-of-the-art MER methods. Keywords: Multimodal emotion recognition, representation subspace mapping, cross-modality attention, unimodal auxiliary loss, fusion

Previously, we introduced a method to generate a multilingual Combinatory Categorial Grammar (CCG) treebank by converting from the Universal Dependencies (UD). However, the method only produces bare CCG derivations without any accompanying semantic representations, which makes it difficult to obtain satisfactory analyses for constructions that involve non-local dependencies, such as control/raising or relative clauses, and limits the general applicability of the treebank. In this work, we present an algorithm that adds semantic representations to existing CCG derivations, in the form of predicate-argument structures. Through hand-crafted rules, we enhance each CCG category with headedness information, with which both local and non-local dependencies can be properly projected. This information is extracted from various sources, including UD, Enhanced UD, and proposition banks. Evaluation of our projected dependencies on the English PropBank and the Universal PropBank 2.0 shows that they can capture most of the semantic dependencies in the target corpora. Further error analysis measures the effectiveness of our algorithm for each language tested, and reveals several issues with the previous method and source data.

Eye movements during reading offer a window into cognitive processes and language comprehension, but the scarcity of reading data with interruptions – which learners frequently encounter in their everyday learning environments – hampers advances in the development of intelligent learning technologies. We introduce InteRead – a novel 50-participant dataset of gaze data recorded during self-paced reading of real-world text. InteRead further offers fine-grained annotations of interruptions interspersed throughout the text as well as resumption lags incurred by these interruptions. Interruptions were triggered automatically once readers reached predefined target words. We validate our dataset by reporting interdisciplinary analyses on different measures of gaze behavior. In line with prior research, our analyses show that the interruptions as well as word length and word frequency effects significantly impact eye movements during reading. We also explore individual differences within our dataset, shedding light on the potential for tailored educational solutions. InteRead is accessible from our datasets web-page: https://www.ife.uni-stuttgart.de/en/llis/research/datasets/.

Information extraction (IE) from text documents is an important NLP task that includes entity, relation, and event extraction. These tasks are often addressed jointly as a graph generation problem, where entities and event triggers represent nodes and where relations and event arguments represent edges. Most existing systems use local classifiers for nodes and edges, trained using cross-entropy loss, and employ inference strategies such as beam search to approximate the optimal graph structure. These approaches typically suffer from exposure bias due to the discrepancy between training and decoding. In this paper, we tackle this problem by casting graph generation as auto-regressive sequence labeling and making its training aware of the decoding procedure by using a differentiable version of beam search. We evaluate the effectiveness of our approach through extensive experiments conducted on the ACE05 and ConLL04 datasets across diverse languages. Our experimental findings affirm that our model outperforms its non-decoding-aware version for all datasets employed. Furthermore, we conduct ablation studies that emphasize the effectiveness of aligning training and inference. Additionally, we introduce a novel quantification of exposure bias within this context, providing valuable insights into the functioning of our model.

In recent years, Multimodal Sentiment Analysis (MSA) leveraging deep learning has demonstrated exceptional performance in a wide range of domains. Its success lies in effectively utilizing information from multiple modalities to analyze sentiments. Despite these advancements, MSA is confronted with two significant challenges. Firstly, each modality often has a surplus of unimportance data, which can overshadow the essential information. Secondly, the crucial cues for sentiment analysis may conflict across different modalities, thereby complicating the analysis process. These issues have a certain impact on the model’s effectiveness in MSA tasks. To address these challenges, this paper introduces a novel method tailored for MSA, termed InfoEnh. This approach utilizes a masking technique as the bottleneck for information filtering, simultaneously maximizing mutual information to retain crucial data. Furthermore, the method integrates all modalities into a common feature space via domain adaptation, which is enhanced by the application of optimal transport. Extensive experiments conducted on two benchmark MSA datasets demonstrate the effectiveness of our proposed approach. Further analyzes indicate significant improvements over the baselines.

Natural Language Inference semantic concepts are central to all aspects of natural language meaning. Portuguese has few NLI-annotated datasets created through automatic translation followed by manual checking. The manual creation of NLI datasets is complex and requires many efforts that are sometimes unavailable. Thus, investments to produce good quality synthetic instances that could be used to train machine learning models for NLI are welcome. This work produced InferBR, an NLI dataset for Portuguese. We relied on a semiautomatic process to generate premises and an automatic process to generate hypotheses. The dataset was manually revised, showing that 97.4% of the sentence pairs had good quality, and nearly 100% of the instances had the correct label assigned. The model trained with InferBR is better at recognizing entailment classes in the other Portuguese datasets than the reverse. Because of its diversity and many unique sentences, InferBR can potentially be further augmented. In addition to the dataset, a key contribution is our proposed generation processes for premises and hypotheses that can easily be adapted to other languages and tasks.

Recently, influence functions present an apparatus for achieving explainability for deep neural models by quantifying the perturbation of individual train instances that might impact a test prediction. Our objectives in this paper are twofold. First we incorporate influence functions as a feedback into the model to improve its performance. Second, in a dataset extension exercise, using influence functions to automatically identify data points that have been initially ‘silver’ annotated by some existing method and need to be cross-checked (and corrected) by annotators to improve the model performance. To meet these objectives, in this paper, we introduce InfFeed, which uses influence functions to compute the influential instances for a target instance. Toward the first objective, we adjust the label of the target instance based on its influencer(s) label. In doing this, InfFeed outperforms the state-of-the-art baselines (including LLMs) by a maximum macro F1-score margin of almost 4% for hate speech classification, 3.5% for stance classification, and 3% for irony and 2% for sarcasm detection. Toward the second objective we show that manually re-annotating only those silver annotated data points in the extension set that have a negative influence can immensely improve the model performance bringing it very close to the scenario where all the data points in the extension set have gold labels. This allows for huge reduction of the number of data points that need to be manually annotated since out of the silver annotated extension dataset, the influence function scheme picks up ~1/1000 points that need manual correction.

The task of inductive knowledge graph completion requires models to learn inference patterns from a training graph, which can then be used to make predictions on a disjoint test graph. Rule-based methods seem like a natural fit for this task, but in practice they significantly underperform state-of-the-art methods based on Graph Neural Networks (GNNs), such as NBFNet. We hypothesise that the underperformance of rule-based methods is due to two factors: (i) implausible entities are not ranked at all and (ii) only the most informative path is taken into account when determining the confidence in a given link prediction answer. To analyse the impact of these factors, we study a number of variants of a rule-based approach, which are specifically aimed at addressing the aforementioned issues. We find that the resulting models can achieve a performance which is close to that of NBFNet. Crucially, the considered variants only use a small fraction of the evidence that NBFNet relies on, which means that they largely keep the interpretability advantage of rule-based methods. Moreover, we show that a further variant, which does look at the full KG, consistently outperforms NBFNet.

Polar questions are common in dialogue and expect exactly one of two answers (yes/no). It is however not uncommon for speakers to bypass these expected choices and answer, for example, “Islands are generally by the sea” to the question: “An island? By the sea?”. While such answers are natural in spoken dialogues, conversational systems still struggle to interpret them. Seminal work to interpret indirect answers were made in recent years—but only for English and with strict question formulations. In this work, we present a new corpus for French and Spanish—IndirectQA —where we mine subtitle data for indirect answers to study the labeling task with six different labels, while broadening polar questions to include also implicit polar questions (statements that trigger a yes/no-answer which are not necessarily formulated as a question). We opted for subtitles since they are a readily available source of conversation in various languages, but also come with peculiarities and challenges which we will discuss. Overall, we provide the first results on French and Spanish. They show that the task is challenging: the baseline accuracy scores drop from 61.43 on English to 44.06 for French and Spanish.

Applications of Natural Language Processing (NLP) in the finance domain have been very popular of late. For financial NLP, (FinNLP) while various datasets exist for widely spoken languages like English and Chinese, datasets are scarce for low resource languages,particularly for Indian languages. In this paper, we address this challenges by presenting IndicFinNLP – a collection of 9 datasets consisting of three tasks relating to FinNLP for three Indian languages. These tasks are Exaggerated Numeral Detection, Sustainability Classification, and ESG Theme Determination of financial texts in Hindi, Bengali, and Telugu. Moreover, we release the datasets under CC BY-NC-SA 4.0 license for the benefit of the research community.