We introduce DataTales, a novel benchmark designed to assess the proficiency of language models in data narration, a task crucial for transforming complex tabular data into accessible narratives. Existing benchmarks often fall short in capturing the requisite analytical complexity for practical applications. DataTales addresses this gap by offering 4.9k financial reports paired with corresponding market data, showcasing the demand for models to create clear narratives and analyze large datasets while understanding specialized terminology in the field. Our findings highlights the significant challenge that language models face in achieving the necessary precision and analytical depth for proficient data narration, suggesting promising avenues for future model development and evaluation methodologies.

We present Multi-expert Prompting, a novel enhancement of ExpertPrompting (Xu et al., 2023), designed to improve the large language model (LLM) generation. Specifically, it guides an LLM to fulfill an input instruction by simulating multiple experts, aggregating their responses, and selecting the best among individual and aggregated responses. This process is performed in a single chain of thoughts through our seven carefully designed subtasks derived from the Nominal Group Technique (Ven and Delbecq, 1974), a well-established decision-making framework. Our evaluations demonstrate that Multi-expert Prompting significantly outperforms ExpertPrompting and comparable baselines in enhancing the truthfulness, factuality, informativeness, and usefulness of responses while reducing toxicity and hurtfulness. It further achieves state-of-the-art truthfulness by outperforming the best baseline by 8.69% with ChatGPT. Multi-expert Prompting is efficient, explainable, and highly adaptable to diverse scenarios, eliminating the need for manual prompt construction.

Image–text models (ITMs) is the prevalent architecture to solve video question–answering tasks, which requires only a few input frames to save huge computational cost compared to video–language models.However, we find existent ITM video question–answering solutions either 1) adopt simplistic and unintentional sampling strategies, which may miss key frames to offer the answer clues; or 2) sample a large number of frames into divided groups, which the computational sources can not accommodate. In this work, we aim at an efficient sampling method towards the few-frame situations.We first summarize a family of prior sampling methods based on question–frame correlation into a unified one, dubbed *Most Implied Frames* (MIF). Through some primary results and analysis, Through analysis, we form a hypothesis that question-aware sampling is not necessary, from which we further propose the other method *Most Dominant Frames* (MDF).Experimental results on four public datasets and three advanced ITMs demonstrate that our proposed strategies can boost the performance for image–text pretrained models, and have a wide application scenario in terms of model architectures and dataset types. Our code is available at https://github.com/declare-lab/Sealinghttps://github.com/declare-lab/Sealing.

Previous zero-shot dialogue state tracking (DST) methods only apply transfer learning, but ignore unlabelled data in the target domain.We transform zero-shot DST into few-shot DST by utilising such unlabelled data via joint and self-training methods. Our method incorporates auxiliary tasks that generate slot types as inverse prompts for main tasks, creating slot values during joint training. Cycle consistency between these two tasks enables the generation and selection of quality samples in unknown target domains for subsequent fine-tuning. This approach also facilitates automatic label creation, thereby optimizing the training and fine-tuning of DST models. We demonstrate this method’s effectiveness on general language models in zero-shot scenarios, improving average joint goal accuracy by 8% across all domains in MultiWOZ.

Large vision-language models (LVLMs) suffer from hallucination, resulting in misalignment between the output textual response and the input visual content. Recent research indicates that the over-reliance on the Large Language Model (LLM) backbone, as one cause of the LVLM hallucination, inherently introduces bias from language priors, leading to insufficient context attention to the visual inputs.We tackle this issue of hallucination by mitigating such over-reliance through preference learning. We propose Vision-guided Direct Preference Optimization (V-DPO) to enhance visual context learning at training time. To interpret the effectiveness and generalizability of V-DPO on different types of training data, we construct a synthetic dataset containing both response- and image-contrast preference pairs, compared against existing human-annotated hallucination samples. Our approach achieves significant improvements compared with baseline methods across various hallucination benchmarks. Our analysis indicates that V-DPO excels in learning from image-contrast preference data, demonstrating its superior ability to elicit and understand nuances of visual context. Our code is publicly available at https://github.com/YuxiXie/V-DPOhttps://github.com/YuxiXie/V-DPO.

Humans perform visual perception at multiple levels, including low-level object recognition and high-level semantic interpretation such as behavior understanding. Subtle differences in low-level details can lead to substantial changes in high-level perception. For example, substituting the shopping bag held by a person with a gun suggests violent behavior, implying criminal or violent activity. Despite significant advancements in various multimodal tasks, Large Visual Language Models (LVLMs) remain unexplored in their capabilities to conduct such multi-level visual perceptions.To investigate the perception gap between LVLMs and humans, we introduce MVP-Bench, the first visual–language benchmark systematically evaluating both low- and high-level visual perception of LVLMs. We construct MVP-Bench across natural and synthetic images to investigate how manipulated content influences model perception. Using MVP-Bench, we diagnose the visual perception of 10 open-source and 2 closed-source LVLMs, showing that high-level perception tasks significantly challenge existing LVLMs. The state-of-the-art GPT-4o only achieves an accuracy of 56% on Yes/No questions, compared with 74% in low-level scenarios. Furthermore, the performance gap between natural and manipulated images indicates that current LVLMs do not generalize in understanding the visual semantics of synthetic images as humans do.

The labor market is changing rapidly, prompting increased interest in the automatic extraction of occupational skills from text. With the advent of English benchmark job description datasets, there is a need for systems that handle their diversity well. We tackle the complexity in occupational skill datasets tasks—combining and leveraging multiple datasets for skill extraction, to identify rarely observed skills within a dataset, and overcoming the scarcity of skills across datasets. In particular, we investigate the retrieval-augmentation of language models, employing an external datastore for retrieving similar skills in a dataset-unifying manner. Our proposed method, Nearest Neighbor Occupational Skill Extraction (NNOSE) effectively leverages multiple datasets by retrieving neighboring skills from other datasets in the datastore. This improves skill extraction without additional fine-tuning. Crucially, we observe a performance gain in predicting infrequent patterns, with substantial gains of up to 30% span-F1 in cross-dataset settings.

Recent developments in Language Models (LMs) have shown their effectiveness in NLP tasks, particularly in knowledge-intensive tasks.However, the mechanisms underlying knowledge storage and memory access within their parameters remain elusive. In this paper, we investigate whether a generative LM (e.g., GPT-2) is able to access its memory sequentially or randomly. Through carefully-designed synthetic tasks, covering the scenarios of full recitation, selective recitation and grounded question answering, we reveal that LMs manage to sequentially access their memory while encountering challenges in randomly accessing memorized content. We find that techniques including recitation and permutation improve the random memory access capability of LMs. Furthermore, by applying this intervention to realistic scenarios of open-domain question answering, we validate that enhancing random access by recitation leads to notable improvements in question answering. The code to reproduce our experiments can be found at https://github.com/sail-sg/lm-random-memory-access.

While large language models have significantly enhanced the effectiveness of discourse relation classifications, it remains unclear whether their comprehension is faithful and reliable. We provide DiSQ, a new method for evaluating the faithfulness of understanding discourse based on question answering. We first employ in-context learning to annotate the reasoning for discourse comprehension, based on the connections among key events within the discourse. Following this, DiSQ interrogates the model with a sequence of questions to assess its grasp of core event relations, its resilience to counterfactual queries, as well as its consistency to its previous responses. then evaluate language models with different architectural designs using DiSQ, finding: (1) DiSQ presents a significant challenge for all models, with the top-performing GPT model attaining only 41% of the ideal performance in PDTB; (2) DiSQ is robust to domain shifts and paraphrase variations; (3) Open-source models generally lag behind their closed-source GPT counterparts, with notable exceptions being those enhanced with chat and code/math features; (4) Our analysis validates the effectiveness of explicitly signalled discourse connectives, the role of contextual information, and the benefits of using historical QA data.

The development of general-domain neural machine translation (NMT) methods has advanced significantly in recent years, but the lack of naturalness and musical constraints in the outputs makes them unable to produce singable lyric translations. This paper bridges the singability quality gap by formalizing lyric translation into a constrained translation problem, converting theoretical guidance and practical techniques from translatology literature to prompt-driven NMT approaches, exploring better adaptation methods, and instantiating them to an English-Chinese lyric translation system. Our model achieves 99.85%, 99.00%, and 95.52% on length accuracy, rhyme accuracy, and word boundary recall. In our subjective evaluation, our model shows a 75% relative enhancement on overall quality, compared against naive fine-tuning (Code available at https://github.com/Sonata165/ControllableLyricTranslation).

A challenge in the Dialogue State Tracking (DST) field is adapting models to new domains without using any supervised data — zero-shot domain adaptation. Parameter-Efficient Transfer Learning (PETL) has the potential to address this problem due to its robustness. However, it has yet to be applied to the zero-shot scenarios, as it is not clear how to apply it unsupervisedly. Our method, Prompter, uses descriptions of target domain slots to generate dynamic prefixes that are concatenated to the key and values at each layer’s self-attention mechanism. This allows for the use of prefix-tuning in zero-shot. Prompter outperforms previous methods on both the MultiWOZ and SGD benchmarks. In generating prefixes, our analyses find that Prompter not only utilizes the semantics of slot descriptions but also how often the slots appear together in conversation. Moreover, Prompter’s gains are due to its improved ability to distinguish ”none”-valued dialogue slots, compared against baselines.

Fact-checking real-world claims often requires collecting multiple pieces of evidence and applying complex multi-step reasoning. In this paper, we present Program-Guided Fact-Checking (ProgramFC), a novel fact-checking model that decomposes complex claims into simpler sub-tasks that can be solved using a shared library of specialized functions. We first leverage the in-context learning ability of large language models to generate reasoning programs to guide the verification process. Afterward, we execute the program by delegating each sub-task to the corresponding sub-task handler. This process makes our model both explanatory and data-efficient, providing clear explanations of its reasoning process and requiring minimal training data. We evaluate ProgramFC on two challenging fact-checking datasets and show that it outperforms seven fact-checking baselines across different settings of evidence availability, with explicit output programs that benefit human debugging. Our codes and data are publicly available at https://github.com/mbzuai-nlp/ProgramFC.

We investigate the potential misuse of modern Large Language Models (LLMs) for generating credible-sounding misinformation and its subsequent impact on information-intensive applications, particularly Open-Domain Question Answering (ODQA) systems. We establish a threat model and simulate potential misuse scenarios, both unintentional and intentional, to assess the extent to which LLMs can be utilized to produce misinformation. Our study reveals that LLMs can act as effective misinformation generators, leading to a significant degradation (up to 87%) in the performance of ODQA systems. Moreover, we uncover disparities in the attributes associated with persuading humans and machines, presenting an obstacle to current human-centric approaches to combat misinformation. To mitigate the harm caused by LLM-generated misinformation, we propose three defense strategies: misinformation detection, vigilant prompting, and reader ensemble. These approaches have demonstrated promising results, albeit with certain associated costs. Lastly, we discuss the practicality of utilizing LLMs as automatic misinformation generators and provide relevant resources and code to facilitate future research in this area.

We introduce ECHo (Event Causality Inference via Human-Centric Reasoning), a diagnostic dataset of event causality inference grounded in visio-linguistic social scenarios. ECHo employs real-world human-centric deductive information building on a television crime drama. ECHo requires the Theory-of-Mind (ToM) ability to understand and reason about social interactions based on multimodal information. Using ECHo, we propose a unified Chain-of-Thought (CoT) framework to assess the reasoning capability of current AI systems. Our ToM-enhanced CoT pipeline accommodates various large foundation models in both zero-shot and few-shot visio-linguistic reasoning. We use this framework to scrutinize recent large foundation models such as InstructGPT and MiniGPT-4 on three diagnostic human-centric tasks. Further analysis demonstrates ECHo as a challenging dataset to expose imperfections and inconsistencies in reasoning. Our data and code are publicly available at [https://github.com/YuxiXie/ECHo](https://github.com/YuxiXie/ECHo).

Annotated data plays a critical role in Natural Language Processing (NLP) in training models and evaluating their performance. Given recent developments in Large Language Models (LLMs), models such as ChatGPT demonstrate zero-shot capability on many text-annotation tasks, comparable with or even exceeding human annotators. Such LLMs can serve as alternatives for manual annotation, due to lower costs and higher scalability. However, limited work has leveraged LLMs as complementary annotators, nor explored how annotation work is best allocated among humans and LLMs to achieve both quality and cost objectives. We propose CoAnnotating, a novel paradigm for Human-LLM co-annotation of unstructured texts at scale. Under this framework, we utilize uncertainty to estimate LLMs’ annotation capability. Our empirical study shows CoAnnotating to be an effective means to allocate work from results on different datasets, with up to 21% performance improvement over random baseline. For code implementation, see https://github.com/SALT-NLP/CoAnnotating.

Current scientific fact-checking benchmarks exhibit several shortcomings, such as biases arising from crowd-sourced claims and an over-reliance on text-based evidence. We present SCITAB, a challenging evaluation dataset consisting of 1.2K expert-verified scientific claims that 1) originate from authentic scientific publications and 2) require compositional reasoning for verification. The claims are paired with evidence-containing scientific tables annotated with labels. Through extensive evaluations, we demonstrate that SCITAB poses a significant challenge to state-of-the-art models, including table-based pretraining models and large language models. All models except GPT-4 achieved performance barely above random guessing. Popular prompting techniques, such as Chain-of-Thought, do not achieve much performance gains on SCITAB. Our analysis uncovers several unique challenges posed by SCITAB, including table grounding, claim ambiguity, and compositional reasoning. Our codes and data are publicly available at https://github.com/XinyuanLu00/SciTab.

We present ACL OCL, a scholarly corpus derived from the ACL Anthology to assist Open scientific research in the Computational Linguistics domain. Integrating and enhancing the previous versions of the ACL Anthology, the ACL OCL contributes metadata, PDF files, citation graphs and additional structured full texts with sections, figures, and links to a large knowledge resource (Semantic Scholar). The ACL OCL spans seven decades, containing 73K papers, alongside 210K figures. We spotlight how ACL OCL applies to observe trends in computational linguistics. By detecting paper topics with a supervised neural model, we note that interest in “Syntax: Tagging, Chunking and Parsing” is waning and “Natural Language Generation” is resurging. Our dataset is available from HuggingFace (https://huggingface.co/datasets/WINGNUS/ACL-OCL).

Improving the quality of academic writing is a meaningful but challenging task. Conventional methods of language refinement focus on narrow, specific linguistic features within isolated sentences, such as grammatical errors and improper word use. We propose a more general task, Academic Writing Formalization (AWF), to improve the overall quality of formal academic writing at the paragraph level. We formulate this language refinement task as a formal text style transfer task which transfers informal-academic text to formal-academic and contribute a large-scale non-parallel dataset, Doolittle, for this purpose. Concurrently, we apply a method named metric-oriented reinforcement learning (MORL) to two large language models (LLM) where we incorporate different levels of automatic feedback into the training process. Our experiments reveal that existing text transfer models and grammatical error correction models address certain aspects of AWF but still have a significant performance gap compared to human performance. Meanwhile, language models fine-tuned with our MORL method exhibit considerably improved performance, rivaling the latest chatbot ChatGPT, but still have a non-negligible gap compared to the ground truth formal-academic texts in Doolittle.

Fact-checking real-world claims often requires intricate, multi-step reasoning due to the absence of direct evidence to support or refute them. However, existing fact-checking systems often lack transparency in their decision-making, making it challenging for users to comprehend their reasoning process. To address this, we propose the Question-guided Multi-hop Fact-Checking (QACheck) system, which guides the model’s reasoning process by asking a series of questions critical for verifying a claim. QACheck has five key modules: a claim verifier, a question generator, a question-answering module, a QA validator, and a reasoner. Users can input a claim into QACheck, which then predicts its veracity and provides a comprehensive report detailing its reasoning process, guided by a sequence of (question, answer) pairs. QACheck also provides the source of evidence supporting each question, fostering a transparent, explainable, and user-friendly fact-checking process.

We present a novel toolkit for controlled summarization of scientific documents, designed for the specific needs of the scientific community. Our system generates summaries based on user preferences, adjusting key attributes specifically of length and keyword inclusion. A distinguishing feature is its ability to manage multiple attributes concurrently, demonstrating Compositional Controllability for Scientific Summarization (CocoSciSum). Benchmarked against the strong Flan-T5 baseline, CocoSciSum exhibits superior performance on both the quality of summaries generated and the control over single and multiple attributes. Moreover, CocoSciSum is a user-centric toolkit, supporting user preferences expressed in natural language instructions, and accommodating diverse input document formats. CocoSciSum is available on GitHub (https://github.com/WING-NUS/SciAssist/tree/CocoSciSum) with an introduction video (https://youtu.be/YC1YDeEjAbQ).

We propose two methods to make unsupervised domain adaptation (UDA) more parameter efficient using adapters – small bottleneck layers interspersed with every layer of the large-scale pre-trained language model (PLM). The first method deconstructs UDA into a two-step process: first by adding a domain adapter to learn domain-invariant information and then by adding a task adapter that uses domain-invariant information to learn task representations in the source domain. The second method jointly learns a supervised classifier while reducing the divergence measure. Compared to strong baselines, our simple methods perform well in natural language inference (MNLI) and the cross-domain sentiment classification task. We even outperform unsupervised domain adaptation methods such as DANN and DSN in sentiment classification, and we are within 0.85% F1 for natural language inference task, by fine-tuning only a fraction of the full model parameters. We release our code at this URL.

Numerous visio-linguistic (V+L) representation learning methods have been developed, yet existing datasets do not adequately evaluate the extent to which they represent visual and linguistic concepts in a unified space. We propose several novel evaluation settings for V+L models, including cross-modal transfer. Furthermore, existing V+L benchmarks often report global accuracy scores on the entire dataset, making it difficult to pinpoint the specific reasoning tasks that models fail and succeed at. We present TraVLR, a synthetic dataset comprising four V+L reasoning tasks. TraVLR’s synthetic nature allows us to constrain its training and testing distributions along task-relevant dimensions, enabling the evaluation of out-of-distribution generalisation. Each example in TraVLR redundantly encodes the scene in two modalities, allowing either to be dropped or added during training or testing without losing relevant information. We compare the performance of four state-of-the-art V+L models, finding that while they perform well on test examples from the same modality, they all fail at cross-modal transfer and have limited success accommodating the addition or deletion of one modality. We release TraVLR as an open challenge for the research community.

With NLP research now quickly being transferred into real-world applications, it is important to be aware of and think through the consequences of our scientific investigation. Such ethical considerations are important in both authoring and reviewing. This tutorial will equip participants with basic guidelines for thinking deeply about ethical issues and review common considerations that recur in NLP research. The methodology is interactive and participatory, including case studies and working in groups. Importantly, the participants will be co-building the tutorial outcomes and will be working to create further tutorial materials to share as public outcomes.

Automatic transfer of text between domains has become popular in recent times. One of its aims is to preserve the semantic content while adapting to the target domain. However, it does not explicitly maintain other attributes between the source and translated text: e.g., text length and descriptiveness. Maintaining constraints in transfer has several downstream applications, including data augmentation and debiasing. We introduce a method for such constrained unsupervised text style transfer by introducing two complementary losses to the generative adversarial network (GAN) family of models. Unlike the competing losses used in GANs, we introduce cooperative losses where the discriminator and the generator cooperate and reduce the same loss. The first is a contrastive loss and the second is a classification loss — aiming to regularize the latent space further and bring similar sentences closer together. We demonstrate that such training retains lexical, syntactic and domain-specific constraints between domains for multiple benchmark datasets, including ones where more than one attribute change. We show that the complementary cooperative losses improve text quality, according to both automated and human evaluation measures.

Due to high data demands of current methods, attention to zero-shot cross-lingual spoken language understanding (SLU) has grown, as such approaches greatly reduce human annotation effort. However, existing models solely rely on shared parameters, which can only perform implicit alignment across languages. We present Global-Local Contrastive Learning Framework (GL-CLeF) to address this shortcoming. Specifically, we employ contrastive learning, leveraging bilingual dictionaries to construct multilingual views of the same utterance, then encourage their representations to be more similar than negative example pairs, which achieves to explicitly align representations of similar sentences across languages. In addition, a key step in GL-CLeF is a proposed Local and Global component, which achieves a fine-grained cross-lingual transfer (i.e., sentence-level Local intent transfer, token-level Local slot transfer, and semantic-level Global transfer across intent and slot). Experiments on MultiATIS++ show that GL-CLeF achieves the best performance and successfully pulls representations of similar sentences across languages closer.

In this paper, we study the problem of knowledge-intensive text-to-SQL, in which domain knowledge is necessary to parse expert questions into SQL queries over domain-specific tables. We formalize this scenario by building a new benchmark KnowSQL consisting of domain-specific questions covering various domains. We then address this problem by representing formulaic knowledge rather than by annotating additional data examples. More concretely, we construct a formulaic knowledge bank as a domain knowledge base and propose a framework (ReGrouP) to leverage this formulaic knowledge during parsing. Experiments using ReGrouP demonstrate a significant 28.2% improvement overall on KnowSQL.

Existing multimodal tasks mostly target at the complete input modality setting, i.e., each modality is either complete or completely missing in both training and test sets. However, the randomly missing situations have still been underexplored. In this paper, we present a novel approach named MM-Align to address the missing-modality inference problem. Concretely, we propose 1) an alignment dynamics learning module based on the theory of optimal transport (OT) for missing data imputation; 2) a denoising training algorithm to enhance the quality of imputation as well as the accuracy of model predictions. Compared with previous generative methods which devote to restoring the missing inputs, MM-Align learns to capture and imitate the alignment dynamics between modality sequences. Results of comprehensive experiments on two multimodal tasks empirically demonstrate that our method can perform more accurate and faster inference and alleviate the overfitting issue under different missing conditions.

Augmentation of task-oriented dialogues has followed standard methods used for plain-text such as back-translation, word-level manipulation, and paraphrasing despite its richly annotated structure. In this work, we introduce an augmentation framework that utilizes belief state annotations to match turns from various dialogues and form new synthetic dialogues in a bottom-up manner. Unlike other augmentation strategies, it operates with as few as five examples. Our augmentation strategy yields significant improvements when both adapting a DST model to a new domain, and when adapting a language model to the DST task, on evaluations with TRADE and TOD-BERT models. Further analysis shows that our model performs better on seen values during training, and it is also more robust to unseen values. We conclude that exploiting belief state annotations enhances dialogue augmentation and results in improved models in n-shot training scenarios.

Modern Natural Language Processing (NLP) models are known to be sensitive to input perturbations and their performance can decrease when applied to real-world, noisy data. However, it is still unclear why models are less robust to some perturbations than others. In this work, we test the hypothesis that the extent to which a model is affected by an unseen textual perturbation (robustness) can be explained by the learnability of the perturbation (defined as how well the model learns to identify the perturbation with a small amount of evidence). We further give a causal justification for the learnability metric. We conduct extensive experiments with four prominent NLP models — TextRNN, BERT, RoBERTa and XLNet — over eight types of textual perturbations on three datasets. We show that a model which is better at identifying a perturbation (higher learnability) becomes worse at ignoring such a perturbation at test time (lower robustness), providing empirical support for our hypothesis.

We model products’ reviews to generate comparative responses consisting of positive and negative experiences regarding the product. Specifically, we generate a single-sentence, comparative response from a given positive and a negative opinion. We contribute the first dataset for this task of Comparative Snippet Generation from contrasting opinions regarding a product, and an analysis of performance of a pre-trained BERT model to generate such snippets.

Logical structure recovery in scientific articles associates text with a semantic section of the article. Although previous work has disregarded the surrounding context of a line, we model this important information by employing line-level attention on top of a transformer-based scientific document processing pipeline. With the addition of loss function engineering and data augmentation techniques with semi-supervised learning, our method improves classification performance by 10% compared to a recent state-of-the-art model. Our parsimonious, text-only method achieves a performance comparable to that of other works that use rich document features such as font and spatial position, using less data without sacrificing performance, resulting in a lightweight training pipeline.

Knowledge-grounded dialog systems need to incorporate smooth transitions among knowledge selected for generating responses, to ensure that dialog flows naturally. For document-grounded dialog systems, the inter- and intra-document knowledge relations can be used to model such conversational flows. We develop a novel Multi-Document Co-Referential Graph (Coref-MDG) to effectively capture the inter-document relationships based on commonsense and similarity and the intra-document co-referential structures of knowledge segments within the grounding documents. We propose CorefDiffs, a Co-referential and Differential flow management method, to linearize the static Coref-MDG into conversational sequence logic. CorefDiffs performs knowledge selection by accounting for contextual graph structures and the knowledge difference sequences. CorefDiffs significantly outperforms the state-of-the-art by 9.5%, 7.4% and 8.2% on three public benchmarks. This demonstrates that the effective modeling of co-reference and knowledge difference for dialog flows are critical for transitions in document-grounded conversation.

Questions of fairness, robustness, and transparency are paramount to address before deploying NLP systems. Central to these concerns is the question of reliability: Can NLP systems reliably treat different demographics fairly and function correctly in diverse and noisy environments? To address this, we argue for the need for reliability testing and contextualize it among existing work on improving accountability. We show how adversarial attacks can be reframed for this goal, via a framework for developing reliability tests. We argue that reliability testing — with an emphasis on interdisciplinary collaboration — will enable rigorous and targeted testing, and aid in the enactment and enforcement of industry standards.

Neural models for automated fact verification have achieved promising results thanks to the availability of large, human-annotated datasets. However, for each new domain that requires fact verification, creating a dataset by manually writing claims and linking them to their supporting evidence is expensive. We develop QACG, a framework for training a robust fact verification model by using automatically generated claims that can be supported, refuted, or unverifiable from evidence from Wikipedia. QACG generates question-answer pairs from the evidence and then converts them into different types of claims. Experiments on the FEVER dataset show that our QACG framework significantly reduces the demand for human-annotated training data. In a zero-shot scenario, QACG improves a RoBERTa model’s F1 from 50% to 77%, equivalent in performance to 2K+ manually-curated examples. Our QACG code is publicly available.

Domain divergence plays a significant role in estimating the performance of a model in new domains. While there is a significant literature on divergence measures, researchers find it hard to choose an appropriate divergence for a given NLP application. We address this shortcoming by both surveying the literature and through an empirical study. We develop a taxonomy of divergence measures consisting of three classes — Information-theoretic, Geometric, and Higher-order measures and identify the relationships between them. Further, to understand the common use-cases of these measures, we recognise three novel applications – 1) Data Selection, 2) Learning Representation, and 3) Decisions in the Wild – and use it to organise our literature. From this, we identify that Information-theoretic measures are prevalent for 1) and 3), and Higher-order measures are more common for 2). To further help researchers choose appropriate measures to predict drop in performance – an important aspect of Decisions in the Wild, we perform correlation analysis spanning 130 domain adaptation scenarios, 3 varied NLP tasks and 12 divergence measures identified from our survey. To calculate these divergences, we consider the current contextual word representations (CWR) and contrast with the older distributed representations. We find that traditional measures over word distributions still serve as strong baselines, while higher-order measures with CWR are effective.

Obtaining training data for multi-hop question answering (QA) is time-consuming and resource-intensive. We explore the possibility to train a well-performed multi-hop QA model without referencing any human-labeled multi-hop question-answer pairs, i.e., unsupervised multi-hop QA. We propose MQA-QG, an unsupervised framework that can generate human-like multi-hop training data from both homogeneous and heterogeneous data sources. MQA-QG generates questions by first selecting/generating relevant information from each data source and then integrating the multiple information to form a multi-hop question. Using only generated training data, we can train a competent multi-hop QA which achieves 61% and 83% of the supervised learning performance for the HybridQA and the HotpotQA dataset, respectively. We also show that pretraining the QA system with the generated data would greatly reduce the demand for human-annotated training data. Our codes are publicly available at https://github.com/teacherpeterpan/Unsupervised-Multi-hop-QA.

The robustness of pretrained language models(PLMs) is generally measured using performance drops on two or more domains. However, we do not yet understand the inherent robustness achieved by contributions from different layers of a PLM. We systematically analyze the robustness of these representations layer by layer from two perspectives. First, we measure the robustness of representations by using domain divergence between two domains. We find that i) Domain variance increases from the lower to the upper layers for vanilla PLMs; ii) Models continuously pretrained on domain-specific data (DAPT)(Gururangan et al., 2020) exhibit more variance than their pretrained PLM counterparts; and that iii) Distilled models (e.g., DistilBERT) also show greater domain variance. Second, we investigate the robustness of representations by analyzing the encoded syntactic and semantic information using diagnostic probes. We find that similar layers have similar amounts of linguistic information for data from an unseen domain.

We introduce a synthetic dialogue generation framework, Velocidapter, which addresses the corpus availability problem for dialogue comprehension. Velocidapter augments datasets by simulating synthetic conversations for a task-oriented dialogue domain, requiring a small amount of bootstrapping work for each new domain. We evaluate the efficacy of our framework on a task-oriented dialogue comprehension dataset, MRCWOZ, which we curate by annotating questions for slots in the restaurant, taxi, and hotel domains of the MultiWOZ 2.2 dataset (Zang et al., 2020). We run experiments within a low-resource setting, where we pretrain a model on SQuAD, fine-tuning it on either a small original data or on the synthetic data generated by our framework. Velocidapter shows significant improvements using both the transformer-based BERTBase and BiDAF as base models. We further show that the framework is easy to use by novice users and conclude that Velocidapter can greatly help training over task-oriented dialogues, especially for low-resourced emerging domains.

We introduce SciWING, an open-source soft-ware toolkit which provides access to state-of-the-art pre-trained models for scientific document processing (SDP) tasks, such as citation string parsing, logical structure recovery and citation intent classification. Compared to other toolkits, SciWING follows a full neural pipeline and provides a Python inter-face for SDP. When needed, SciWING provides fine-grained control for rapid experimentation with different models by swapping and stacking different modules. Transfer learning from general and scientific documents specific pre-trained transformers (i.e., BERT, SciBERT, etc.) can be performed. SciWING incorporates ready-to-use web and terminal-based applications and demonstrations to aid adoption and development. The toolkit is available from http://sciwing.io and the demos are available at http://rebrand.ly/sciwing-demo.

Recent question generation (QG) approaches often utilize the sequence-to-sequence framework (Seq2Seq) to optimize the log likelihood of ground-truth questions using teacher forcing. However, this training objective is inconsistent with actual question quality, which is often reflected by certain global properties such as whether the question can be answered by the document. As such, we directly optimize for QG-specific objectives via reinforcement learning to improve question quality. We design three different rewards that target to improve the fluency, relevance, and answerability of generated questions. We conduct both automatic and human evaluations in addition to thorough analysis to explore the effect of each QG-specific reward. We find that optimizing on question-specific rewards generally leads to better performance in automatic evaluation metrics. However, only the rewards that correlate well with human judgement (e.g., relevance) lead to real improvement in question quality. Optimizing for the others, especially answerability, introduces incorrect bias to the model, resulting in poorer question quality. The code is publicly available at https://github.com/YuxiXie/RL-for-Question-Generation.

Research into the area of multiparty dialog has grown considerably over recent years. We present the Molweni dataset, a machine reading comprehension (MRC) dataset with discourse structure built over multiparty dialog. Molweni’s source samples from the Ubuntu Chat Corpus, including 10,000 dialogs comprising 88,303 utterances. We annotate 30,066 questions on this corpus, including both answerable and unanswerable questions. Molweni also uniquely contributes discourse dependency annotations in a modified Segmented Discourse Representation Theory (SDRT; Asher et al., 2016) style for all of its multiparty dialogs, contributing large-scale (78,245 annotated discourse relations) data to bear on the task of multiparty dialog discourse parsing. Our experiments show that Molweni is a challenging dataset for current MRC models: BERT-wwm, a current, strong SQuAD 2.0 performer, achieves only 67.7% F1 on Molweni’s questions, a 20+% significant drop as compared against its SQuAD 2.0 performance.

We introduce a deep learning model to learn the set of enumerated job skills associated with a job description. In our analysis of a large-scale government job portal mycareersfuture.sg, we observe that as much as 65% of job descriptions miss describing a significant number of relevant skills. Our model addresses this task from the perspective of an extreme multi-label classification (XMLC) problem, where descriptions are the evidence for the binary relevance of thousands of individual skills. Building upon the current state-of-the-art language modeling approaches such as BERT, we show our XMLC method improves on an existing baseline solution by over 9% and 7% absolute improvements in terms of recall and normalized discounted cumulative gain. We further show that our approach effectively addresses the missing skills problem, and helps in recovering relevant skills that were missed out in the job postings by taking into account the structured semantic representation of skills and their co-occurrences through a Correlation Aware Bootstrapping process. We further show that our approach, to ensure the BERT-XMLC model accounts for structured semantic representation of skills and their co-occurrences through a Correlation Aware Bootstrapping process, effectively addresses the missing skills problem, and helps in recovering relevant skills that were missed out in the job postings. To facilitate future research and replication of our work, we have made the dataset and the implementation of our model publicly available.

The curse of knowledge can impede communication between experts and laymen. We propose a new task of expertise style transfer and contribute a manually annotated dataset with the goal of alleviating such cognitive biases. Solving this task not only simplifies the professional language, but also improves the accuracy and expertise level of laymen descriptions using simple words. This is a challenging task, unaddressed in previous work, as it requires the models to have expert intelligence in order to modify text with a deep understanding of domain knowledge and structures. We establish the benchmark performance of five state-of-the-art models for style transfer and text simplification. The results demonstrate a significant gap between machine and human performance. We also discuss the challenges of automatic evaluation, to provide insights into future research directions. The dataset is publicly available at https://srhthu.github.io/expertise-style-transfer/.

This paper proposes the problem of Deep Question Generation (DQG), which aims to generate complex questions that require reasoning over multiple pieces of information about the input passage. In order to capture the global structure of the document and facilitate reasoning, we propose a novel framework that first constructs a semantic-level graph for the input document and then encodes the semantic graph by introducing an attention-based GGNN (Att-GGNN). Afterward, we fuse the document-level and graph-level representations to perform joint training of content selection and question decoding. On the HotpotQA deep-question centric dataset, our model greatly improves performance over questions requiring reasoning over multiple facts, leading to state-of-the-art performance. The code is publicly available at https://github.com/WING-NUS/SG-Deep-Question-Generation.

Training on only perfect Standard English corpora predisposes pre-trained neural networks to discriminate against minorities from non-standard linguistic backgrounds (e.g., African American Vernacular English, Colloquial Singapore English, etc.). We perturb the inflectional morphology of words to craft plausible and semantically similar adversarial examples that expose these biases in popular NLP models, e.g., BERT and Transformer, and show that adversarially fine-tuning them for a single epoch significantly improves robustness without sacrificing performance on clean data.

Inflectional variation is a common feature of World Englishes such as Colloquial Singapore English and African American Vernacular English. Although comprehension by human readers is usually unimpaired by non-standard inflections, current NLP systems are not yet robust. We propose Base-Inflection Encoding (BITE), a method to tokenize English text by reducing inflected words to their base forms before reinjecting the grammatical information as special symbols. Fine-tuning pretrained NLP models for downstream tasks using our encoding defends against inflectional adversaries while maintaining performance on clean data. Models using BITE generalize better to dialects with non-standard inflections without explicit training and translation models converge faster when trained with BITE. Finally, we show that our encoding improves the vocabulary efficiency of popular data-driven subword tokenizers. Since there has been no prior work on quantitatively evaluating vocabulary efficiency, we propose metrics to do so.

In existing sophisticated text-to-SQL models, schema linking is often considered as a simple, minor component, belying its importance. By providing a schema linking corpus based on the Spider text-to-SQL dataset, we systematically study the role of schema linking. We also build a simple BERT-based baseline, called Schema-Linking SQL (SLSQL) to perform a data-driven study. We find when schema linking is done well, SLSQL demonstrates good performance on Spider despite its structural simplicity. Many remaining errors are attributable to corpus noise. This suggests schema linking is the crux for the current text-to-SQL task. Our analytic studies provide insights on the characteristics of schema linking for future developments of text-to-SQL tasks.

Graph Convolutional Networks (GCNs) are a class of spectral clustering techniques that leverage localized convolution filters to perform supervised classification directly on graphical structures. While such methods model nodes’ local pairwise importance, they lack the capability to model global importance relative to other nodes of the graph. This causes such models to miss critical information in tasks where global ranking is a key component for the task, such as in keyphrase extraction. We address this shortcoming by allowing the proper incorporation of global information into the GCN family of models through the use of scaled node weights. In the context of keyphrase extraction, incorporating global random walk scores obtained from TextRank boosts performance significantly. With our proposed method, we achieve state-of-the-art results, bettering a strong baseline by an absolute 2% increase in F1 score.

Users participate in online discussion forums to learn from others and share their knowledge with the community. They often start a thread with a question or by sharing their new findings on a certain topic. We find that, unlike Community Question Answering, where questions are mostly factoid based, the threads in a forum are often open-ended (e.g., asking for recommendations from others) without a single correct answer. In this paper, we address the task of identifying helpful posts in a forum thread to help users comprehend long running discussion threads, which often contain repetitive or irrelevant posts. We propose a recurrent neural network based architecture to model (i) the relevance of a post regarding the original post starting the thread and (ii) the novelty it brings to the discussion, compared to the previous posts in the thread. Experimental results on different types of online forum datasets show that our model significantly outperforms the state-of-the-art neural network models for text classification.

Sentiment ambiguous lexicons refer to words where their polarity depends strongly on con- text. As such, when the context is absent, their translations or their embedded sentence ends up (incorrectly) being dependent on the training data. While neural machine translation (NMT) has achieved great progress in recent years, most systems aim to produce one single correct translation for a given source sentence. We investigate the translation variation in two sentiment scenarios. We perform experiments to study the preservation of sentiment during translation with three different methods that we propose. We conducted tests with both sentiment and non-sentiment bearing contexts to examine the effectiveness of our methods. We show that NMT can generate both positive- and negative-valent translations of a source sentence, based on a given input sentiment label. Empirical evaluations show that our valence-sensitive embedding (VSE) method significantly outperforms a sequence-to-sequence (seq2seq) baseline, both in terms of BLEU score and ambiguous word translation accuracy in test, given non-sentiment bearing contexts.

Datasets are integral artifacts of empirical scientific research. However, due to natural language variation, their recognition can be difficult and even when identified, can often be inconsistently referred across and within publications. We report our approach to the Coleridge Initiative’s Rich Context Competition, which tasks participants with identifying dataset surface forms (dataset mention extraction) and associating the extracted mention to its referred dataset (dataset classification). In this work, we propose various neural baselines and evaluate these model on one-plus and zero-shot classification scenarios. We further explore various joint learning approaches - exploring the synergy between the tasks - and report the issues with such techniques.

Existing solutions to task-oriented dialogue systems follow pipeline designs which introduces architectural complexity and fragility. We propose a novel, holistic, extendable framework based on a single sequence-to-sequence (seq2seq) model which can be optimized with supervised or reinforcement learning. A key contribution is that we design text spans named belief spans to track dialogue believes, allowing task-oriented dialogue systems to be modeled in a seq2seq way. Based on this, we propose a simplistic Two Stage CopyNet instantiation which emonstrates good scalability: significantly reducing model complexity in terms of number of parameters and training time by a magnitude. It significantly outperforms state-of-the-art pipeline-based methods on large datasets and retains a satisfactory entity match rate on out-of-vocabulary (OOV) cases where pipeline-designed competitors totally fail.

Identifying emergent research trends is a key issue for both primary researchers as well as secondary research managers. Such processes can uncover the historical development of an area, and yield insight on developing topics. We propose an embedded trend detection framework for this task which incorporates our bijunctive hypothesis that important phrases are written by important authors within a field and vice versa. By ranking both author and phrase information in a multigraph, our method jointly determines key phrases and authoritative authors. We represent this intermediate output as phrasal embeddings, and feed this to a recurrent neural network (RNN) to compute trend scores that identify research trends. Over two large datasets of scientific articles, we demonstrate that our approach successfully detects past trends from the field, outperforming baselines based solely on text centrality or citation.

The Association of Computational Linguistic’s Anthology is the open source archive, and the main source for computational linguistics and natural language processing’s scientific literature. The ACL Anthology is currently maintained exclusively by community volunteers and has to be available and up-to-date at all times. We first discuss the current, open source approach used to achieve this, and then discuss how the planned use of Docker images will improve the Anthology’s long-term stability. This change will make it easier for researchers to utilize Anthology data for experimentation. We believe the ACL community can directly benefit from the extension-friendly architecture of the Anthology. We end by issuing an open challenge of reviewer matching we encourage the community to rally towards.

We systematically confirm that instructors are strongly influenced by the user interface presentation of Massive Online Open Course (MOOC) discussion forums. In a large scale dataset, we conclusively show that instructor interventions exhibit strong position bias, as measured by the position where the thread appeared on the user interface at the time of intervention. We measure and remove this bias, enabling unbiased statistical modelling and evaluation. We show that our de-biased classifier improves predicting interventions over the state-of-the-art on courses with sufficient number of interventions by 8.2% in F1 and 24.4% in recall on average.

With Health 2.0, patients and caregivers increasingly seek information regarding possible drug side effects during their medical treatments in online health communities. These are helpful platforms for non-professional medical opinions, yet pose risk of being unreliable in quality and insufficient in quantity to cover the wide range of potential drug reactions. Existing approaches which analyze such user-generated content in online forums heavily rely on feature engineering of both documents and users, and often overlook the relationships between posts within a common discussion thread. Inspired by recent advancements, we propose a neural architecture that models the textual content of user-generated documents and user experiences in online communities to predict side effects during treatment. Experimental results show that our proposed architecture outperforms baseline models.

We describe an end-to-end pipeline processing approach for SemEval 2017’s Task 10 to extract keyphrases and their relations from scientific publications. We jointly identify and classify keyphrases by modeling the subtasks as sequential labeling. Our system utilizes standard, surface-level features along with the adjacent word features, and performs conditional decoding on whole text to extract keyphrases. We focus only on the identification and typing of keyphrases (Subtasks A and B, together referred as extraction), but provide an end-to-end system inclusive of keyphrase relation identification (Subtask C) for completeness. Our top performing configuration achieves an F₁ of 0.27 for the end-to-end keyphrase extraction and relation identification scenario on the final test data, and compares on par to other top ranked systems for keyphrase extraction. Our system outperforms other techniques that do not employ global decoding and hence do not account for dependencies between keyphrases. We believe this is crucial for keyphrase classification in the given context of scientific document mining.

Patients turn to Online Health Communities not only for information on specific conditions but also for emotional support. Previous research has indicated that the progression of emotional status can be studied through the linguistic patterns of an individual’s posts. We analyze a real-world dataset from the Mental Health section of HealthBoards.com. Estimated from the word usages in their posts, we find that the emotional progress across patients vary widely. We study the problem of predicting a patient’s emotional status in the future from her past posts and we propose a Recurrent Neural Network (RNN) based architecture to address it. We find that the future emotional status can be predicted with reasonable accuracy given her historical posts and participation features. Our evaluation results demonstrate the efficacy of our proposed architecture, by outperforming state-of-the-art approaches with over 0.13 reduction in Mean Absolute Error.

Word embeddings are now ubiquitous forms of word representation in natural language processing. There have been applications of word embeddings for monolingual word sense disambiguation (WSD) in English, but few comparisons have been done. This paper attempts to bridge that gap by examining popular embeddings for the task of monolingual English WSD. Our simplified method leads to comparable state-of-the-art performance without expensive retraining. Cross-Lingual WSD – where the word senses of a word in a source language come from a separate target translation language – can also assist in language learning; for example, when providing translations of target vocabulary for learners. Thus we have also applied word embeddings to the novel task of cross-lingual WSD for Chinese and provide a public dataset for further benchmarking. We have also experimented with using word embeddings for LSTM networks and found surprisingly that a basic LSTM network does not work well. We discuss the ramifications of this outcome.

The ACL Anthology is a digital archive of conference and journal papers in natural language processing and computational linguistics. Its primary purpose is to serve as a reference repository of research results, but we believe that it can also be an object of study and a platform for research in its own right. We describe an enriched and standardized reference corpus derived from the ACL Anthology that can be used for research in scholarly document processing. This corpus, which we call the ACL Anthology Reference Corpus (ACL ARC), brings together the recent activities of a number of research groups around the world. Our goal is to make the corpus widely available, and to encourage other researchers to use it as a standard testbed for experiments in both bibliographic and bibliometric research.

We describe ParsCit, a freely available, open-source implementation of a reference string parsing package. At the core of ParsCit is a trained conditional random field (CRF) model used to label the token sequences in the reference string. A heuristic model wraps this core with added functionality to identify reference strings from a plain text file, and to retrieve the citation contexts. The package comes with utilities to run it as a web service or as a standalone utility. We compare ParsCit on three distinct reference string datasets and show that it compares well with other previously published work.