NAACL 2022 Papers

Content-based collaborative filtering (CCF) predicts user-item interactions based on both users’ interaction history and items’ content information. Recently, pre-trained language models (PLM) have been used to extract high-quality item encodings for CCF. However, it is resource-intensive to train a PLM-based CCF model in an end-to-end (E2E) manner, since optimization involves back-propagating through every content encoding within a given user interaction sequence. To tackle this issue, we propose GRAM (GRadient Accumulation for Multi-modality in CCF), which exploits the fact that a given item often appears multiple times within a batch of interaction histories. Specifically, Single-step GRAM aggregates each item encoding’s gradients for back-propagation, with theoretic equivalence to the standard E2E training. As an extension of Single-step GRAM, we propose Multi-step GRAM, which increases the gradient update latency, achieving a further speedup with drastically less GPU memory. GRAM significantly improves training efficiency (up to 146x) on five datasets from two task domains of Knowledge Tracing and News Recommendation. Our code is available at https://github.com/yoonseok312/GRAM.

Dense retrieval approaches can overcome the lexical gap and lead to significantly improved search results. However, they require large amounts of training data which is not available for most domains. As shown in previous work (Thakur et al., 2021b), the performance of dense retrievers severely degrades under a domain shift. This limits the usage of dense retrieval approaches to only a few domains with large training datasets. In this paper, we propose the novel unsupervised domain adaptation method Generative Pseudo Labeling (GPL), which combines a query generator with pseudo labeling from a cross-encoder. On six representative domain-specialized datasets, we find the proposed GPL can outperform an out-of-the-box state-of-the-art dense retrieval approach by up to 9.3 points nDCG@10. GPL requires less (unlabeled) data from the target domain and is more robust in its training than previous methods. We further investigate the role of six recent pre-training methods in the scenario of domain adaptation for retrieval tasks, where only three could yield improved results. The best approach, TSDAE (Wang et al., 2021) can be combined with GPL, yielding another average improvement of 1.4 points nDCG@10 across the six tasks. The code and the models are available at https://github.com/UKPLab/gpl.

Document Information Extraction (DIE) has attracted increasing attention due to its various advanced applications in the real world. Although recent literature has already achieved competitive results, these approaches usually fail when dealing with complex documents with noisy OCR results or mutative layouts. This paper proposes Generative Multi-modal Network (GMN) for real-world scenarios to address these problems, which is a robust multi-modal generation method without predefined label categories. With the carefully designed spatial encoder and modal-aware mask module, GMN can deal with complex documents that are hard to serialized into sequential order. Moreover, GMN tolerates errors in OCR results and requires no character-level annotation, which is vital because fine-grained annotation of numerous documents is laborious and even requires annotators with specialized domain knowledge. Extensive experiments show that GMN achieves new state-of-the-art performance on several public DIE datasets and surpasses other methods by a large margin, especially in realistic scenes.

The past several years have witnessed Variational Auto-Encoder’s superiority in various text generation tasks. However, due to the sequential nature of the text, auto-regressive decoders tend to ignore latent variables and then reduce to simple language models, known as the KL vanishing problem, which would further deteriorate when VAE is combined with Transformer-based structures. To ameliorate this problem, we propose Della, a novel variational Transformer framework. Della learns a series of layer-wise latent variables with each inferred from those of lower layers and tightly coupled with the hidden states by low-rank tensor product. In this way, Della forces these posterior latent variables to be fused deeply with the whole computation path and hence incorporate more information. We theoretically demonstrate that our method can be regarded as entangling latent variables to avoid posterior information decrease through layers, enabling Della to get higher non-zero KL values even without any annealing or thresholding tricks. Experiments on four unconditional and three conditional generation tasks show that Della could better alleviate KL vanishing and improve both quality and diversity compared to several strong baselines.

In this paper, we formulate system combination for grammatical error correction (GEC) as a simple machine learning task: binary classification. We demonstrate that with the right problem formulation, a simple logistic regression algorithm can be highly effective for combining GEC models. Our method successfully increases the F0.5 score from the highest base GEC system by 4.2 points on the CoNLL-2014 test set and 7.2 points on the BEA-2019 test set. Furthermore, our method outperforms the state of the art by 4.0 points on the BEA-2019 test set, 1.2 points on the CoNLL-2014 test set with original annotation, and 3.4 points on the CoNLL-2014 test set with alternative annotation. We also show that our system combination generates better corrections with higher F0.5 scores than the conventional ensemble.

Due to the dialogue characteristics of unstructured contexts and multi-parties with first-person perspective, many successful text summarization works have failed when dealing with dialogue summarization. In dialogue summarization task, the input dialogue is usually spoken style with ellipsis and co-references but the output summaries are more formal and complete. Therefore, the dialogue summarization model should be able to complete the ellipsis content and co-reference information and then produce a suitable summary accordingly. However, the current state-of-the-art models pay more attention on the topic or structure of summary, rather than the consistency of dialogue summary with its input dialogue context, which may suffer from the personal and logical inconsistency problem. In this paper, we propose a new model, named ReWriteSum, to tackle this problem. Firstly, an utterance rewriter is conducted to complete the ellipsis content of dialogue content and then obtain the rewriting utterances. Then, the co-reference data augmentation mechanism is utilized to replace the referential person name with its specific name to enhance the personal information. Finally, the rewriting utterances and the co-reference replacement data are used in the standard BART model. Experimental results on both SAMSum and DialSum datasets show that our ReWriteSum significantly outperforms baseline models, in terms of both metric-based and human evaluations. Further analysis on multi-speakers also shows that ReWriteSum can obtain relatively higher improvement with more speakers, validating the correctness and property of ReWriteSum.

Understanding, modelling and predicting human risky decision-making is challenging due to intrinsic individual differences and irrationality. Fuzzy trace theory (FTT) is a powerful paradigm that explains human decision-making by incorporating gists, i.e., fuzzy representations of information which capture only its quintessential meaning. Inspired by Broniatowski and Reyna’s FTT cognitive model, we propose a computational framework which combines the effects of the underlying semantics and sentiments on text-based decision-making. In particular, we introduce Category-2-Vector to learn categorical gists and categorical sentiments, and demonstrate how our computational model can be optimised to predict risky decision-making in groups and individuals.

Cross-lingual transfer (CLT) is of various applications. However, labeled cross-lingual corpus is expensive or even inaccessible, especially in the fields where labels are private, such as diagnostic results of symptoms in medicine and user profiles in business. Nevertheless, there are off-the-shelf models in these sensitive fields. Instead of pursuing the original labels, a workaround for CLT is to transfer knowledge from the off-the-shelf models without labels. To this end, we define a novel CLT problem named FreeTransfer-X that aims to achieve knowledge transfer from the off-the-shelf models in rich-resource languages. To address the problem, we propose a 2-step knowledge distillation (KD, Hinton et al., 2015) framework based on multilingual pre-trained language models (mPLM). The significant improvement over strong neural machine translation (NMT) baselines demonstrates the effectiveness of the proposed method. In addition to reducing annotation cost and protecting private labels, the proposed method is compatible with different networks and easy to be deployed. Finally, a range of analyses indicate the great potential of the proposed method.

We present a novel approach incorporating transformer-based language models into infectious disease modelling. Text-derived features are quantified by tracking high-density clusters of sentence-level representations of Reddit posts within specific US states’ COVID-19 subreddits. We benchmark these clustered embedding features against features extracted from other high-quality datasets. In a threshold-classification task, we show that they outperform all other feature types at predicting upward trend signals, a significant result for infectious disease modelling in areas where epidemiological data is unreliable. Subsequently, in a time-series forecasting task, we fully utilise the predictive power of the caseload and compare the relative strengths of using different supplementary datasets as covariate feature sets in a transformer-based time-series model.

This paper presents a conversational AI platform called Flowstorm. Flowstorm is an open-source SaaS project suitable for creating, running, and analyzing conversational applications. Thanks to the fast and fully automated build process, the dialogues created within the platform can be executed in seconds. Furthermore, we propose a novel dialogue architecture that uses a combination of tree structures with generative models. The tree structures are also used for training NLU models suitable for specific dialogue scenarios. However, the generative models are globally used across applications and extend the functionality of the dialogue trees. Moreover, the platform functionality benefits from out-of-the-box components, such as the one responsible for extracting data from utterances or working with crawled data. Additionally, it can be extended using a custom code directly in the platform. One of the essential features of the platform is the possibility to reuse the created assets across applications. There is a library of prepared assets where each developer can contribute. All of the features are available through a user-friendly visual editor.

It is challenging to train a good intent classifier for a task-oriented dialogue system with only a few annotations. Recent studies have shown that fine-tuning pre-trained language models with a small set of labeled utterances from public benchmarks in a supervised manner is extremely helpful. However, we find that supervised pre-training yields an anisotropic feature space, which may suppress the expressive power of the semantic representations. Inspired by recent research in isotropization, we propose to improve supervised pre-training by regularizing the feature space towards isotropy. We propose two regularizers based on contrastive learning and correlation matrix respectively, and demonstrate their effectiveness through extensive experiments. Our main finding is that it is promising to regularize supervised pre-training with isotropization to further improve the performance of few-shot intent detection. The source code can be found at https://github.com/fanolabs/isoIntentBert-main.

Modern image captioning models are usually trained with text similarity objectives. However, since reference captions in public datasets often describe the most salient common objects, models trained with the text similarity objectives tend to ignore specific and detailed aspects of an image that distinguish it from others. Towards more descriptive and distinctive caption generation, we propose to use CLIP, a multimodal encoder trained on huge image-text pairs from the web, to calculate multi-modal similarity and use it as a reward function. We also propose a simple finetuning strategy of CLIP text encoder to improve grammar that does not require extra text annotation. This completely eliminates the need for reference captions during the reward computation. To comprehensively evaluate descriptive captions, we introduce FineCapEval, a new dataset for caption evaluation with fine-grained criteria: overall, background, object, relations. In our experiments on text-to-image retrieval and FineCapEval, the proposed CLIP-guided model generates more distinctive captions than the CIDEroptimized model. We also show that our unsupervised grammar finetuning of the CLIP text encoder alleviates the degeneration problem of the naive CLIP reward. Lastly, we show human analysis where the annotators strongly prefer CLIP reward to CIDEr and MLE objectives on diverse criteria.

Pre-trained language models have shown successful progress in many text understanding benchmarks. This work explores the capability of these models to predict actionable plans in real-world environments. Given a text instruction, we show that language priors encoded in pre-trained models allow us to infer fine-grained subgoal sequences. In contrast to recent methods which make strong assumptions about subgoal supervision, our experiments show that language models can infer detailed subgoal sequences from few training sequences without any fine-tuning. We further propose a simple strategy to re-rank language model predictions based on interaction and feedback from the environment. Combined with pre-trained navigation and visual reasoning components, our approach demonstrates competitive performance on subgoal prediction and task completion in the ALFRED benchmark compared to prior methods that assume more subgoal supervision.

Large language models can perform semantic parsing with little training data, when prompted with in-context examples. It has been shown that this can be improved by formulating the problem as paraphrasing into canonical utterances, which casts the underlying meaning representation into a controlled natural language-like representation. Intuitively, such models can more easily output canonical utterances as they are closer to the natural language used for pre-training. Recently, models also pre-trained on code, like OpenAI Codex, have risen in prominence. For semantic parsing tasks where we map natural language into code, such models may prove more adept at it. In this paper, we test this hypothesis and find that Codex performs better on such tasks than equivalent GPT-3 models. We evaluate on Overnight and SMCalFlow and find that unlike GPT-3, Codex performs similarly when targeting meaning representations directly, perhaps because meaning representations are structured similar to code in these datasets.

Self-rationalization models that predict task labels and generate free-text elaborations for their predictions could enable more intuitive interaction with NLP systems. These models are, however, currently trained with a large amount of human-written free-text explanations for each task which hinders their broader usage. We propose to study a more realistic setting of self-rationalization using few training examples. We present FEB—a standardized collection of four existing English-language datasets and associated metrics. We identify the right prompting approach by extensively exploring natural language prompts on FEB. Then, by using this prompt and scaling the model size, we demonstrate that making progress on few-shot self-rationalization is possible. We show there is still ample room for improvement in this task: the average plausibility of generated explanations assessed by human annotators is at most 51% (with GPT-3), while plausibility of human explanations is 76%. We hope that FEB and our proposed approach will spur the community to take on the few-shot self-rationalization challenge.

We present FREDo, a few-shot document-level relation extraction (FSDLRE) benchmark. As opposed to existing benchmarks which are built on sentence-level relation extraction corpora, we argue that document-level corpora provide more realism, particularly regarding none-of-the-above (NOTA) distributions. Therefore, we propose a set of FSDLRE tasks and construct a benchmark based on two existing supervised learning data sets, DocRED and sciERC. We adapt the state-of-the-art sentence-level method MNAV to the document-level and develop it further for improved domain adaptation. We find FSDLRE to be a challenging setting with interesting new characteristics such as the ability to sample NOTA instances from the support set. The data, code, and trained models are available online (https://github.com/nicpopovic/FREDo).

Machine Learning (ML) systems are getting increasingly popular, and drive more and more applications and services in our daily life. Thishas led to growing concerns over user privacy, since human interaction data typically needs to be transmitted to the cloud in order to trainand improve such systems. Federated learning (FL) has recently emerged as a method for training ML models on edge devices using sensitive user data and is seen as a way to mitigate concerns over data privacy. However, since ML models are most commonly trained with label supervision, we need a way to extract labels on edge to make FL viable. In this work, we propose a strategy for training FL models using positive and negative user feedback. We also design a novel framework to study different noise patterns in user feedback, and explore how well standard noise-robust objectives can help mitigate this noise when training models in a federated setting. We evaluate our proposed training setup through detailed experiments on two text classification datasets and analyze the effects of varying levels of user reliability and feedback noise on model performance. We show that our method improves substantially over a self-training baseline, achieving performance closer to models trained with full supervision.

Increasing concerns and regulations about data privacy and sparsity necessitate the study of privacy-preserving, decentralized learning methods for natural language processing (NLP) tasks. Federated learning (FL) provides promising approaches for a large number of clients (e.g., personal devices or organizations) to collaboratively learn a shared global model to benefit all clients while allowing users to keep their data locally. Despite interest in studying FL methods for NLP tasks, a systematic comparison and analysis is lacking in the literature. Herein, we present the FedNLP, a benchmarking framework for evaluating federated learning methods on four different task formulations: text classification, sequence tagging, question answering, and seq2seq. We propose a universal interface between Transformer-based language models (e.g., BERT, BART) and FL methods (e.g., FedAvg, FedOPT, etc.) under various non-IID partitioning strategies. Our extensive experiments with FedNLP provide empirical comparisons between FL methods and help us better understand the inherent challenges of this direction. The comprehensive analysis points to intriguing and exciting future research aimed at developing FL methods for NLP tasks.

Avoiding to rely on dataset artifacts to predict hate speech is at the cornerstone of robust and fair hate speech detection. In this paper we critically analyze lexical biases in hate speech detection via a cross-platform study, disentangling various types of spurious and authentic artifacts and analyzing their impact on out-of-distribution fairness and robustness. We experiment with existing approaches and propose simple yet surprisingly effective data-centric baselines. Our results on English data across four platforms show that distinct spurious artifacts require different treatments to ultimately attain both robustness and fairness in hate speech detection. To encourage research in this direction, we release all baseline models and the code to compute artifacts, pointing it out as a complementary and necessary addition to the data statements practice.