COLING 2024 论文列表

Situated conversations, which refer to visual information as visual question answering (VQA), often contain ambiguities caused by reliance on directive information. This problem is exacerbated because some languages, such as Japanese, often omit subjective or objective terms. Such ambiguities in questions are often clarified by the contexts in conversational situations, such as joint attention with a user or user gaze information. In this study, we propose the Gaze-grounded VQA dataset (GazeVQA) that clarifies ambiguous questions using gaze information by focusing on a clarification process complemented by gaze information. We also propose a method that utilizes gaze target estimation results to improve the accuracy of GazeVQA tasks. Our experimental results showed that the proposed method improved the performance in some cases of a VQA system on GazeVQA and identified some typical problems of GazeVQA tasks that need to be improved.

We introduce a frustratingly simple, highly efficient, and surprisingly effective decoding method, termed Frustratingly Simple Decoding (FSD), for neural text generation. The idea behind FSD is straightforward: We construct an anti-language model (anti-LM) based on previously generated text, which is employed to penalize the future generation of repetitive content. The anti-LM can be implemented as simple as an n-gram language model or a vectorized variant. In this way, FSD incurs no additional model parameters and negligible computational overhead (FSD can be as fast as greedy search). Despite its simplicity, FSD is surprisingly effective and generalizes across different datasets, models, and languages. Extensive experiments show that FSD outperforms established strong baselines in terms of generation quality, decoding speed, and universality.

With an auxiliary corpus (non-target speaker corpus) for model pre-training, Text-to-Speech (TTS) methods can generate high-quality speech with a limited target speaker corpus. However, this approach comes with expensive training costs. To overcome the challenge, a high-quality TTS method is proposed, significantly reducing training costs while maintaining the naturalness of synthesized speech. In this paper, we propose an auxiliary corpus compression algorithm that reduces the training cost while the naturalness of the synthesized speech is not significantly degraded. We then use the compressed corpus to pre-train the proposed TTS model CMDTTS, which fuses phoneme and word multi-level prosody modeling components and denoises the generated mel-spectrograms using denoising diffusion probabilistic models (DDPMs). In addition, a fine-tuning step that the conditional generative adversarial network (cGAN) is introduced to embed the target speaker feature and improve speech quality using the target speaker corpus. Experiments are conducted on Chinese and English single speaker’s corpora, and the results show that the method effectively balances the model training speed and the synthesized speech quality and outperforms the current models.

Transformer language models (LMs) are fundamental to NLP research methodologies and applications in various languages. However, developing such models specifically for the Russian language has received little attention. This paper introduces a collection of 13 Russian Transformer LMs, which spans encoder (ruBERT, ruRoBERTa, ruELECTRA), decoder (ruGPT-3), and encoder-decoder (ruT5, FRED-T5) architectures. We provide a report on the model architecture design and pretraining, and the results of evaluating their generalization abilities on Russian language understanding and generation datasets and benchmarks. By pretraining and releasing these specialized Transformer LMs, we aim to broaden the scope of the NLP research directions and enable the development of industrial solutions for the Russian language.

Radiology report analysis provides valuable information that can aid with public health initiatives, and has been attracting increasing attention from the research community. In this work, we present a novel insight that the structure of a radiology report (namely, the Findings and Impression sections) offers different views of a radiology scan. Based on this intuition, we further propose a co-training approach, where two machine learning models are built upon the Findings and Impression sections, respectively, and use each other’s information to boost performance with massive unlabeled data in a semi-supervised manner. We conducted experiments in a public health surveillance study, and results show that our co-training approach is able to improve performance using the dual views and surpass competing supervised and semi-supervised methods.

Document-level text simplification, a task that combines single-document summarization and intra-sentence simplification, has garnered significant attention. However, studies have primarily focused on languages such as English and German, leaving Japanese and similar languages underexplored because of a scarcity of linguistic resources. In this study, we devised JADOS, the first Japanese document-level text simplification dataset based on newspaper articles and Wikipedia. Our dataset focuses on simplification, to enhance readability by reducing the number of sentences and tokens in a document. We conducted investigations using our dataset. Firstly, we analyzed the characteristics of Japanese simplification by comparing it across different domains and with English counterparts. Moreover, we experimentally evaluated the performances of text summarization methods, transformer-based text simplification models, and large language models. In terms of D-SARI scores, the transformer-based models performed best across all domains. Finally, we manually evaluated several model outputs and target articles, demonstrating the need for document-level text simplification models in Japanese.

Integer Linear Programming (ILP) has been proposed as a formalism for encoding precise structural and semantic constraints for Natural Language Inference (NLI). However, traditional ILP frameworks are non-differentiable, posing critical challenges for the integration of continuous language representations based on deep learning. In this paper, we introduce a novel approach, named Diff-Comb Explainer, a neuro-symbolic architecture for explanation-based NLI based on Differentiable BlackBox Combinatorial Solvers (DBCS). Differently from existing neuro-symbolic solvers, Diff-Comb Explainer does not necessitate a continuous relaxation of the semantic constraints, enabling a direct, more precise, and efficient incorporation of neural representations into the ILP formulation. Our experiments demonstrate that Diff-Comb Explainer achieves superior performance when compared to conventional ILP solvers, neuro-symbolic black-box solvers, and Transformer-based encoders. Moreover, a deeper analysis reveals that Diff-Comb Explainer can significantly improve the precision, consistency, and faithfulness of the constructed explanations, opening new opportunities for research on neuro-symbolic architectures for explainable and transparent NLI in complex domains.

The proliferation of open knowledge graphs has led to a surge in scholarly research on the topic over the past decade. This paper presents a bibliometric analysis of the scholarly literature on open knowledge graphs published between 2013 and 2023. The study aims to identify the trends, patterns, and impact of research in this field, as well as the key topics and research questions that have emerged. The work uses bibliometric techniques to analyze a sample of 4445 scholarly articles retrieved from Scopus. The findings reveal an ever-increasing number of publications on open knowledge graphs published every year, particularly in developed countries (+50 per year). These outputs are published in highly-referred scholarly journals and conferences. The study identifies three main research themes: (1) knowledge graph construction and enrichment, (2) evaluation and reuse, and (3) fusion of knowledge graphs into NLP systems. Within these themes, the study identifies specific tasks that have received considerable attention, including entity linking, knowledge graph embedding, and graph neural networks.

User-generated data sources have gained significance in uncovering Adverse Drug Reactions (ADRs), with an increasing number of discussions occurring in the digital world. However, the existing clinical corpora predominantly revolve around scientific articles in English. This work presents a multilingual corpus of texts concerning ADRs gathered from diverse sources, including patient fora, social media, and clinical reports in German, French, and Japanese. Our corpus contains annotations covering 12 entity types, four attribute types, and 13 relation types. It contributes to the development of real-world multilingual language models for healthcare. We provide statistics to highlight certain challenges associated with the corpus and conduct preliminary experiments resulting in strong baselines for extracting entities and relations between these entities, both within and across languages.

In this study, we present a novel historical Chinese dataset for named entity recognition, entity linking, coreference and entity relations. We use data from Chinese newspapers from 1872 to 1949 and multilingual bibliographic resources from the same period. The period and the language are the main strength of the present work, offering a resource which covers different styles and language uses, as well as the largest historical Chinese NER dataset with manual annotations from this transitional period. After detailing the selection and annotation process, we present the very first results that can be obtained from this dataset. Texts and annotations are freely downloadable from the GitHub repository.

When assessing second language proficiency (L2), evaluation of spontaneous speech performance is crucial. This paper presents a corpus of spontaneous L2 English speech, focusing on the speech performance of B1 and B2 proficiency speakers. Two hundred and sixty university students were recorded during a speaking task as part of a French national certificate in English. This task entailed a 10-minute role-play among 2 or 3 candidates, arguing about a controversial topic, in order to reach a negotiated compromise. Each student’s performance was evaluated by two experts, categorizing them into B2, B1 or below B1 speaking proficiency levels. Automatic diarization, transcription, and alignment at the word level were performed on the recorded conversations, in order to analyse lexical stress realisation in polysyllabic plain words of B1 and B2 proficiency students. Results showed that only 35.4% of the 6,350 targeted words had stress detected on the expected syllable, revealing a common stress shift to the final syllable. Besides a substantial inter-speaker variability (0% to 68.4%), B2 speakers demonstrated a slightly higher stress accuracy (36%) compared to B1 speakers (29.6%). Those with accurate stress placement utilized F0 and intensity to make syllable prominence, while speakers with lower accuracy tended to lengthen words on their last syllables, with minimal changes in other dimensions.

We present the first comprehensive set of guidelines for German Abstract Meaning Representation (Deutsche AMR, DeAMR) along with an annotated corpus of 400 DeAMR. Taking English AMR (EnAMR) as our starting point, we propose significant adaptations to faithfully represent the structure and semantics of German, focusing particularly on verb frames, compound words, and modality. We validate our annotation through inter-annotator agreement and further evaluate our corpus with a comparison of structural divergences between EnAMR and DeAMR on parallel sentences, replicating previous work that finds both cases of cross-lingual structural alignment and cases of meaningful linguistic divergence. Finally, we fine-tune state-of-the-art multi-lingual and cross-lingual AMR parsers on our corpus and find that, while our small corpus is insufficient to produce quality output, there is a need to continue develop and evaluate against gold non-English AMR data.

We develop novel annotation guidelines for sentence-level subjectivity detection, which are not limited to language-specific cues. We use our guidelines to collect NewsSD-ENG, a corpus of 638 objective and 411 subjective sentences extracted from English news articles on controversial topics. Our corpus paves the way for subjectivity detection in English and across other languages without relying on language-specific tools, such as lexicons or machine translation. We evaluate state-of-the-art multilingual transformer-based models on the task in mono-, multi-, and cross-language settings. For this purpose, we re-annotate an existing Italian corpus. We observe that models trained in the multilingual setting achieve the best performance on the task.

Manufacturing specifications are documents entailing different techniques, processes, and components involved in manufacturing. There is a growing demand for named entity recognition (NER) resources and techniques for manufacturing-specific named entities, with the development of smart manufacturing. In this paper, we introduce a corpus of Chinese manufacturing specifications, named MS-NERC, including 4,424 sentences and 16,383 entities. We also propose an entity recognizer named Trainable State Transducer (TST), which is initialized with a finite state transducer describing the morphological patterns of entities. It can directly recognize entities based on prior morphological knowledge without training. Experimental results show that TST achieves an overall 82.05% F1 score for morphological-specific entities in zero-shot. TST can be improved through training, the result of which outperforms neural methods in few-shot and rich-resource. We believe that our corpus and model will be valuable resources for NER research not only in manufacturing but also in other low-resource domains.

All state-of-the-art coreference resolution (CR) models involve finetuning a pretrained language model. Whether the superior performance of one CR model over another is due to the choice of language model or other factors, such as the task-specific architecture, is difficult or impossible to determine due to lack of a standardized experimental setup. To resolve this ambiguity, we systematically evaluate five CR models and control for certain design decisions including the pretrained language model used by each. When controlling for language model size, encoder-based CR models outperform more recent decoder-based models in terms of both accuracy and inference speed. Surprisingly, among encoder-based CR models, more recent models are not always more accurate, and the oldest CR model that we test generalizes the best to out-of-domain textual genres. We conclude that controlling for the choice of language model reduces most, but not all, of the increase in F1 score reported in the past five years.

Large Language Models (LLMs) have been developed without a theoretical framework, yet we posit that evaluating and improving LLMs will benefit from the development of theoretical frameworks that enable comparison of the structures of human language and the model of language built up by LLMs through the processing of text. In service of this goal, we develop the Construction Grammar Schematicity (“CoGS”) corpus of 10 distinct English constructions, where the constructions vary with respect to schematicity, or in other words the level to which constructional slots require specific, fixed lexical items, or can be filled with a variety of elements that fulfill a particular semantic role of the slot. Our corpus constructions are carefully curated to range from substantive, frozen constructions (e.g., Let-alone) to entirely schematic constructions (e.g., Resultative). The corpus was collected to allow us to probe LLMs for constructional information at varying levels of abstraction. We present our own probing experiments using this corpus, which clearly demonstrate that even the largest LLMs are limited to more substantive constructions and do not exhibit recognition of the similarity of purely schematic constructions. We publicly release our dataset, prompts, and associated model responses.

Pictographs have been found to improve patient comprehension of medical information or instructions. However, tools to produce pictograph representations from natural language are still scarce. In this contribution we describe a system that automatically translates French speech into pictographs to enable diagnostic interviews in emergency settings, thereby providing a tool to overcome the language barrier or provide support in Augmentative and Alternative Communication (AAC) contexts. Our approach is based on a semantic gloss that serves as pivot between spontaneous language and pictographs, with medical concepts represented using the UMLS ontology. In this study we evaluate different available pre-trained models fine-tuned on artificial data to translate French into this semantic gloss. On unseen data collected in real settings, consisting of questions and instructions by physicians, the best model achieves an F0.5 score of 86.7. A complementary human evaluation of the semantic glosses differing from the reference shows that 71% of these would be usable to transmit the intended meaning. Finally, a human evaluation of the pictograph sequences derived from the gloss reveals very few additions, omissions or order issues (<3%), suggesting that the gloss as designed is well suited as a pivot for translation into pictographs.

In this paper we describe a computational model of Latvian morphology that provides a formal structure for Latvian word form inflection and has been implemented in software for generation, analysis and lemmatization of Latvian word forms. The work was motivated by the need for a NLP inflection model that can cover all the complexity of Latvian language and explicitly enumerate and handle the many exceptions to the general Latvian inflection principles. This is an evolution of earlier work, extending the initial proof of concept model to properly cover Latvian language. We provide a set of morphological paradigms that differ from current linguistic tradition, a set of systematic stem changes and combine it with an extensive lexicon that includes paradigm information and structured morphological attributes for 118 000 lexemes. This model has been applied on both dictionary and corpora data, demonstrating that it provides a good coverage for modern Latvian literary language. We also consider that there is a good potential to extend this also to the related Latgalian language.

This paper explores grammaticization of deverbal prepositions by a computational approach based on corpus data. Deverbal prepositions are words or phrases that are derived from a verb and that behave as a preposition such as “regarding” and “according to”. Linguistic studies have revealed important aspects of grammaticization of deverbal prepositions. This paper augments them by methods for measuring the degree of grammaticization of deverbal prepositions based on non-contextualized or contextualized word vectors. Experiments show that the methods correlate well with human judgements (as high as 0.69 in Spearman’s rank correlation coefficient). Using the best-performing method, this paper further shows that the methods support previous findings in linguistics including (i) Deverbal prepositions are marginal in terms of prepositionality; and (ii) The process where verbs are grammaticized into prepositions is gradual. As a pilot study, it also conducts a diachronic analysis of grammaticization of deverbal preposition.