Masahiro Kaneko

Daisuke Oba, Masahiro Kaneko, Danushka Bollegala

Despite their impressive performance in a wide range of NLP tasks, Large Language Models (LLMs) have been reported to encode worrying-levels of gender biases. Prior work has proposed debiasing methods that require human labelled examples, data augmentation and fine-tuning of LLMs, which are computationally costly. Moreover, one might not even have access to the model parameters for performing debiasing such as in the case of closed LLMs such as GPT-4. To address this challenge, we propose bias suppression that prevents biased generations of LLMs by simply providing textual preambles constructed from manually designed templates and real-world statistics, without accessing to model parameters. We show that, using CrowsPairs dataset, our textual preambles covering counterfactual statements can suppress gender biases in English LLMs such as LLaMA2. Moreover, we find that gender-neutral descriptions of gender-biased objects can also suppress their gender biases. Moreover, we show that bias suppression has acceptable adverse effect on downstream task performance with HellaSwag and COPA.

Masahiro Kaneko, Danushka Bollegala, Timothy Baldwin

Recent studies have demonstrated that large language models (LLMs) have ethical-related problems such as social biases, lack of moral reasoning, and generation of offensive content. The existing evaluation metrics and methods to address these ethical challenges use datasets intentionally created by instructing humans to create instances including ethical problems. Therefore, the data does not reflect prompts that users actually provide when utilizing LLM services in everyday contexts. This may not lead to the development of safe LLMs that can address ethical challenges arising in real-world applications. In this paper, we create Eagle datasets extracted from real interactions between ChatGPT and users that exhibit social biases, toxicity, and immoral problems. Our experiments show that Eagle captures complementary aspects, not covered by existing datasets proposed for evaluation and mitigation of such ethical challenges. Our code is publicly available at https://huggingface.co/datasets/MasahiroKaneko/eagle.

Masahiro Kaneko, Masato Mita, Shun Kiyono et al.

This paper investigates how to effectively incorporate a pre-trained masked language model (MLM), such as BERT, into an encoder-decoder (EncDec) model for grammatical error correction (GEC). The answer to this question is not as straightforward as one might expect because the previous common methods for incorporating a MLM into an EncDec model have potential drawbacks when applied to GEC. For example, the distribution of the inputs to a GEC model can be considerably different (erroneous, clumsy, etc.) from that of the corpora used for pre-training MLMs; however, this issue is not addressed in the previous methods. Our experiments show that our proposed method, where we first fine-tune a MLM with a given GEC corpus and then use the output of the fine-tuned MLM as additional features in the GEC model, maximizes the benefit of the MLM. The best-performing model achieves state-of-the-art performances on the BEA-2019 and CoNLL-2014 benchmarks. Our code is publicly available at: https://github.com/kanekomasahiro/bert-gec.

Masahiro Kaneko, Zeerak Talat, Timothy Baldwin

Iterative jailbreak methods that repeatedly rewrite and input prompts into large language models (LLMs) to induce harmful outputs -- using the model's previous responses to guide each new iteration -- have been found to be a highly effective attack strategy. Despite being an effective attack strategy against LLMs and their safety mechanisms, existing defenses do not proactively disrupt this dynamic trial-and-error cycle. In this study, we propose a novel framework that dynamically updates its defense strategy through online learning in response to each new prompt from iterative jailbreak methods. Leveraging the distinctions between harmful jailbreak-generated prompts and typical harmless prompts, we introduce a reinforcement learning-based approach that optimizes prompts to ensure appropriate responses for harmless tasks while explicitly rejecting harmful prompts. Additionally, to curb overfitting to the narrow band of partial input rewrites explored during an attack, we introduce Past-Direction Gradient Damping (PDGD). Experiments conducted on three LLMs show that our approach significantly outperforms five existing defense methods against five iterative jailbreak methods. Moreover, our results indicate that our prompt optimization strategy simultaneously enhances response quality for harmless tasks.

Masahiro Kaneko, Danushka Bollegala, Timothy Baldwin

The output tendencies of Pre-trained Language Models (PLM) vary markedly before and after Fine-Tuning (FT) due to the updates to the model parameters. These divergences in output tendencies result in a gap in the social biases of PLMs. For example, there exits a low correlation between intrinsic bias scores of a PLM and its extrinsic bias scores under FT-based debiasing methods. Additionally, applying FT-based debiasing methods to a PLM leads to a decline in performance in downstream tasks. On the other hand, PLMs trained on large datasets can learn without parameter updates via In-Context Learning (ICL) using prompts. ICL induces smaller changes to PLMs compared to FT-based debiasing methods. Therefore, we hypothesize that the gap observed in pre-trained and FT models does not hold true for debiasing methods that use ICL. In this study, we demonstrate that ICL-based debiasing methods show a higher correlation between intrinsic and extrinsic bias scores compared to FT-based methods. Moreover, the performance degradation due to debiasing is also lower in the ICL case compared to that in the FT case.

Seiichiro Kondo, Kengo Hotate, Masahiro Kaneko et al.

Neural machine translation (NMT) has recently gained widespread attention because of its high translation accuracy. However, it shows poor performance in the translation of long sentences, which is a major issue in low-resource languages. It is assumed that this issue is caused by insufficient number of long sentences in the training data. Therefore, this study proposes a simple data augmentation method to handle long sentences. In this method, we use only the given parallel corpora as the training data and generate long sentences by concatenating two sentences. Based on the experimental results, we confirm improvements in long sentence translation by the proposed data augmentation method, despite its simplicity. Moreover, the translation quality is further improved by the proposed method, when combined with back-translation.

Aomi Koyama, Kengo Hotate, Masahiro Kaneko et al.

Grammatical error correction (GEC) suffers from a lack of sufficient parallel data. Therefore, GEC studies have developed various methods to generate pseudo data, which comprise pairs of grammatical and artificially produced ungrammatical sentences. Currently, a mainstream approach to generate pseudo data is back-translation (BT). Most previous GEC studies using BT have employed the same architecture for both GEC and BT models. However, GEC models have different correction tendencies depending on their architectures. Thus, in this study, we compare the correction tendencies of the GEC models trained on pseudo data generated by different BT models, namely, Transformer, CNN, and LSTM. The results confirm that the correction tendencies for each error type are different for every BT model. Additionally, we examine the correction tendencies when using a combination of pseudo data generated by different BT models. As a result, we find that the combination of different BT models improves or interpolates the F_0.5 scores of each error type compared with that of single BT models with different seeds.

Raj Dabre, Aizhan Imankulova, Masahiro Kaneko et al.

Parallel corpora are indispensable for training neural machine translation (NMT) models, and parallel corpora for most language pairs do not exist or are scarce. In such cases, pivot language NMT can be helpful where a pivot language is used such that there exist parallel corpora between the source and pivot and pivot and target languages. Naturally, the quality of pivot language translation is more inferior to what could be achieved with a direct parallel corpus of a reasonable size for that pair. In a real-time simultaneous translation setting, the quality of pivot language translation deteriorates even further given that the model has to output translations the moment a few source words become available. To solve this issue, we propose multi-pivot translation and apply it to a simultaneous translation setting involving pivot languages. Our approach involves simultaneously translating a source language into multiple pivots, which are then simultaneously translated together into the target language by leveraging multi-source NMT. Our experiments in a low-resource setting using the N-way parallel UN corpus for Arabic to English NMT via French and Spanish as pivots reveals that in a simultaneous pivot NMT setting, using two pivot languages can lead to an improvement of up to 5.8 BLEU.

Masahiro Kaneko, Danushka Bollegala

Word embeddings trained on large corpora have shown to encode high levels of unfair discriminatory gender, racial, religious and ethnic biases. In contrast, human-written dictionaries describe the meanings of words in a concise, objective and an unbiased manner. We propose a method for debiasing pre-trained word embeddings using dictionaries, without requiring access to the original training resources or any knowledge regarding the word embedding algorithms used. Unlike prior work, our proposed method does not require the types of biases to be pre-defined in the form of word lists, and learns the constraints that must be satisfied by unbiased word embeddings automatically from dictionary definitions of the words. Specifically, we learn an encoder to generate a debiased version of an input word embedding such that it (a) retains the semantics of the pre-trained word embeddings, (b) agrees with the unbiased definition of the word according to the dictionary, and (c) remains orthogonal to the vector space spanned by any biased basis vectors in the pre-trained word embedding space. Experimental results on standard benchmark datasets show that the proposed method can accurately remove unfair biases encoded in pre-trained word embeddings, while preserving useful semantics.

Masahiro Kaneko, Danushka Bollegala

In comparison to the numerous debiasing methods proposed for the static non-contextualised word embeddings, the discriminative biases in contextualised embeddings have received relatively little attention. We propose a fine-tuning method that can be applied at token- or sentence-levels to debias pre-trained contextualised embeddings. Our proposed method can be applied to any pre-trained contextualised embedding model, without requiring to retrain those models. Using gender bias as an illustrative example, we then conduct a systematic study using several state-of-the-art (SoTA) contextualised representations on multiple benchmark datasets to evaluate the level of biases encoded in different contextualised embeddings before and after debiasing using the proposed method. We find that applying token-level debiasing for all tokens and across all layers of a contextualised embedding model produces the best performance. Interestingly, we observe that there is a trade-off between creating an accurate vs. unbiased contextualised embedding model, and different contextualised embedding models respond differently to this trade-off.

Masahiro Kaneko, Danushka Bollegala

Prior work investigating the geometry of pre-trained word embeddings have shown that word embeddings to be distributed in a narrow cone and by centering and projecting using principal component vectors one can increase the accuracy of a given set of pre-trained word embeddings. However, theoretically, this post-processing step is equivalent to applying a linear autoencoder to minimise the squared l2 reconstruction error. This result contradicts prior work (Mu and Viswanath, 2018) that proposed to remove the top principal components from pre-trained embeddings. We experimentally verify our theoretical claims and show that retaining the top principal components is indeed useful for improving pre-trained word embeddings, without requiring access to additional linguistic resources or labelled data.

Masato Mita, Shun Kiyono, Masahiro Kaneko et al.

Existing approaches for grammatical error correction (GEC) largely rely on supervised learning with manually created GEC datasets. However, there has been little focus on verifying and ensuring the quality of the datasets, and on how lower-quality data might affect GEC performance. We indeed found that there is a non-negligible amount of "noise" where errors were inappropriately edited or left uncorrected. To address this, we designed a self-refinement method where the key idea is to denoise these datasets by leveraging the prediction consistency of existing models, and outperformed strong denoising baseline methods. We further applied task-specific techniques and achieved state-of-the-art performance on the CoNLL-2014, JFLEG, and BEA-2019 benchmarks. We then analyzed the effect of the proposed denoising method, and found that our approach leads to improved coverage of corrections and facilitated fluency edits which are reflected in higher recall and overall performance.

Aizhan Imankulova, Masahiro Kaneko, Tosho Hirasawa et al.

Simultaneous translation involves translating a sentence before the speaker's utterance is completed in order to realize real-time understanding in multiple languages. This task is significantly more challenging than the general full sentence translation because of the shortage of input information during decoding. To alleviate this shortage, we propose multimodal simultaneous neural machine translation (MSNMT), which leverages visual information as an additional modality. Our experiments with the Multi30k dataset showed that MSNMT significantly outperforms its text-only counterpart in more timely translation situations with low latency. Furthermore, we verified the importance of visual information during decoding by performing an adversarial evaluation of MSNMT, where we studied how models behaved with incongruent input modality and analyzed the effect of different word order between source and target languages.

Masahiro Kaneko, Danushka Bollegala

Word embeddings learnt from massive text collections have demonstrated significant levels of discriminative biases such as gender, racial or ethnic biases, which in turn bias the down-stream NLP applications that use those word embeddings. Taking gender-bias as a working example, we propose a debiasing method that preserves non-discriminative gender-related information, while removing stereotypical discriminative gender biases from pre-trained word embeddings. Specifically, we consider four types of information: \emph{feminine}, \emph{masculine}, \emph{gender-neutral} and \emph{stereotypical}, which represent the relationship between gender vs. bias, and propose a debiasing method that (a) preserves the gender-related information in feminine and masculine words, (b) preserves the neutrality in gender-neutral words, and (c) removes the biases from stereotypical words. Experimental results on several previously proposed benchmark datasets show that our proposed method can debias pre-trained word embeddings better than existing SoTA methods proposed for debiasing word embeddings while preserving gender-related but non-discriminative information.

Masahiro Kaneko, Mamoru Komachi

It is known that a deep neural network model pre-trained with large-scale data greatly improves the accuracy of various tasks, especially when there are resource constraints. However, the information needed to solve a given task can vary, and simply using the output of the final layer is not necessarily sufficient. Moreover, to our knowledge, exploiting large language representation models to detect grammatical errors has not yet been studied. In this work, we investigate the effect of utilizing information not only from the final layer but also from intermediate layers of a pre-trained language representation model to detect grammatical errors. We propose a multi-head multi-layer attention model that determines the appropriate layers in Bidirectional Encoder Representation from Transformers (BERT). The proposed method achieved the best scores on three datasets for grammatical error detection tasks, outperforming the current state-of-the-art method by 6.0 points on FCE, 8.2 points on CoNLL14, and 12.2 points on JFLEG in terms of F_0.5. We also demonstrate that by using multi-head multi-layer attention, our model can exploit a broader range of information for each token in a sentence than a model that uses only the final layer's information.

Masato Mita, Tomoya Mizumoto, Masahiro Kaneko et al.

This study explores the necessity of performing cross-corpora evaluation for grammatical error correction (GEC) models. GEC models have been previously evaluated based on a single commonly applied corpus: the CoNLL-2014 benchmark. However, the evaluation remains incomplete because the task difficulty varies depending on the test corpus and conditions such as the proficiency levels of the writers and essay topics. To overcome this limitation, we evaluate the performance of several GEC models, including NMT-based (LSTM, CNN, and transformer) and an SMT-based model, against various learner corpora (CoNLL-2013, CoNLL-2014, FCE, JFLEG, ICNALE, and KJ). Evaluation results reveal that the models' rankings considerably vary depending on the corpus, indicating that single-corpus evaluation is insufficient for GEC models.