Croatia

Masked language modeling (MLM) plays a key role in pretraining large language models. But the MLM objective is often dominated by high-frequency words that are sub-optimal for learning factual knowledge. In this work, we propose an approach for influencing MLM pretraining in a way that can improve language model performance on a variety of knowledge-intensive tasks. We force the language model to prioritize informative words in a fully unsupervised way. Experiments demonstrate that the proposed approach can significantly improve the performance of pretrained language models on tasks such as factual recall, question answering, sentiment analysis, and natural language inference in a closed-book setting.

EACL 2023

Unsupervised Improvement of Factual Knowledge in Language Models

knowledge enhanced language model

pretraining strategy

unsupervised

poster

### Welcome to the 17th Conference of the European Chapter of the Association for Computational Linguistics (EACL) 
Continuing its mission of expanding and involving the science community of all European countries, EACL had selected the Ukrainian community for the 16th EACL, which took place online due to the COVID pandemic. Unfortunately, the ongoing war made the organisation in Kyiv impossible. Considering the importance of physical interaction among researchers, especially after the restrictions imposed by the COVID pandemic, in addition to an online mode, the conference will be held in Dubrovnik, Croatia, from 2 to 6 of May, 2023. The original aim of strengthening the connection with the Ukrainian community will not change as our program will feature a dedicated session and a workshop to highlight work on Ukrainian language technologies. As the flagship European conference in the field of computational linguistics, EACL welcomes European and international researchers covering a broad spectrum of research areas that are concerned with computational approaches to natural language.

You need to register for the conference in order to access this site. Please visit https://2023.eacl.org/registration for more information.

As the flagship European conference in the field of computational linguistics, EACL welcomes European and international researchers covering a broad spectrum of research areas that are concerned with computational approaches to natural language.

Despite the strong performance of current NLP models, they can be brittle against adversarial attacks. To enable effective learning against adversarial inputs, we introduce the use of rationale models that can explicitly learn to ignore attack tokens. We find that the rationale models can successfully ignore over 90% of attack tokens. This approach leads to consistent sizable improvements (∼10%) over baseline models in robustness on three datasets for both BERT and RoBERTa, and also reliably outperforms data augmentation with adversarial examples alone. In many cases, we find that our method is able to close the gap between model performance on a clean test set and an attacked test set and hence reduce the effect of adversarial attacks.

Learning to Ignore Adversarial Attacks

Understanding sentence semantics requires an interpretation of the main information from a concrete context. To investigate how individual word contributes to sentence semantics, we propose a perturbation method for unsupervised semantic analysis. We next re-examine SOTA sentence embedding models’ ability to capture the main semantics of a sentence by developing a new evaluation metric to adapt sentence compression datasets for automatic evaluation. Results on three datasets show that unsupervised discourse relation recognition can serve as a general inference task that can more effectively aggregate information to essential contents than several SOTA unsupervised sentence embedding models.


How do Words Contribute to Sentence Semantics? Revisiting Sentence Embeddings with a Perturbation Method

Building natural language inference (NLI) benchmarks that are both challenging for modern techniques, and free from shortcut biases is difficult. Chief among these biases is "single sentence label leakage," where annotator-introduced spurious correlations yield datasets where the logical relation between (premise, hypothesis) pairs can be accurately predicted from only a single sentence, something that should in principle be impossible. We demonstrate that despite efforts to reduce this leakage, it persists in modern datasets that have been introduced since its 2018 discovery. To enable future amelioration efforts, introduce a novel model-driven technique, the progressive evaluation of cluster outliers (PECO) which enables both the objective measurement of leakage, and the automated detection of subpopulations in the data which maximally exhibit it.

PECO: Examining Single Sentence Label Leakage in Natural Language Inference Datasets through Progressive Evaluation of Cluster Outliers

Biomedical data and benchmarks are highly valuable yet very limited in low-resource languages other than English, such as Vietnamese. In this paper, we use a state-of-the-art translation model in English-Vietnamese to translate and produce both pretrained and supervised data in the biomedical domains. Thanks to such large-scale translation, we introduce ViPubmedT5, a pretrained Encoder-Decoder Transformer model trained on 20 million translated abstracts from the high-quality public PubMed corpus. ViPubMedT5 demonstrates state-of-the-art results on two different biomedical benchmarks in summarization and acronym disambiguation. Further, we release ViMedNLI - a new NLP task in Vietnamese translated from MedNLI using the recently public En-vi translation model and carefully refined by human experts, with evaluations of existing methods against ViPubmedT5.

Enriching Biomedical Knowledge for Low-resource Language Through Large-scale Translation

Montague Grammar seeks to use a sentence’s syntactic structure to model its compositional semantics. However, semantic representations in the deep learning era are no longer symbolic and not easily modeled by syntax. How can we explore whether syntax serves as a strong causal model of vectorial compositional semantics? We introduce Syntactic Neural Module Nets (SynNaMoNs) to enable the computation of sentence embeddings as modeled by a sentence’s syntactic structure. We show how SynNaMoNs can be trained via distillation from a sentence transformer used as an embedding teacher model. Treating the optimal distillation loss as a measure of ‘distillability’, we discuss how the distillability of a transformer’s sentence embeddings to a SynNaMoN measures how well a sentence transformer models compositional semantics. Empirically, we explore 5 sentence embedding models, 3 syntax module internal architectures, and the intuitive generalization behavior of a determiner phrase module.

Syntax-guided Neural Module Distillation to Probe Compositionality in Sentence Embeddings

The event extraction task typically consists of event detection and event argument extraction. Most previous work models these two subtasks with shared representation by multiple classification tasks or a unified generative approach. In this paper, we revisit this pattern and propose to use independent encoders to model event detection and event argument extraction, respectively, and use the output of event detection to construct the input of event argument extraction. In addition, we use token-level features to precisely control the fusion between two encoders to achieve joint bridging training rather than directly reusing representations between different tasks. Through a series of careful experiments, we demonstrate the importance of avoiding feature interference of different tasks and the importance of joint bridging training. We achieved competitive results on standard benchmarks (ACE05-E, ACE05-E+, and ERE-EN) and established a solid baseline.

A Hybrid Detection and Generation Framework with Separate Encoders for Event Extraction

Emotion-Cause Pair Extraction (ECPE) task aims to pair all emotions and corresponding causes in documents.
ECPE is an important task for developing human-like responses.
However, previous ECPE research is conducted based on news articles, which has different characteristics compared to dialogues.
To address this issue, we propose a Pair-Relationship Guided Mixture-of-Experts (PRG-MoE) model, which considers dialogue features (e.g., speaker information).
PRG-MoE automatically learns relationship between utterances and advises a gating network to incorporate dialogue features in the evaluation, yielding substantial performance improvement.
We employ a new ECPE dataset, which is an English dialogue dataset, with more emotion-cause pairs in documents than news articles.
We also propose Cause Type Classification that classifies emotion-cause pairs according to the types of the cause of a detected emotion.
For reproducing the results, we make available all our code and data (https://github.com/jdjin3000/PRG-MoE).

Conversational Emotion-Cause Pair Extraction with Guided Mixture of Experts

End-to-end neural models for conversational AI often assume that a response can be generated by considering only the knowledge acquired by the model during training. Document-oriented conversational models make a similar assumption by conditioning the input on the document and assuming that any other knowledge is captured in the model’s weights. However, a conversation may refer to external knowledge sources. In this work, we present EKo-DoC, an architecture for document-oriented conversations with access to external knowledge: we assume that a conversation is centered around a topic document and that external knowledge is needed to produce responses. EKo-DoC includes a dense passage retriever, a re-ranker, and a response generation model. We train the model end-to-end by using silver labels for the retrieval and re-ranking components that we automatically acquire from the attention signals of the response generation model. We demonstrate with automatic and human evaluations that incorporating external knowledge improves response generation in document-oriented conversations. Our architecture achieves new state-of-the-art results on the Wizard of Wikipedia dataset, outperforming a competitive baseline by 10.3% in Recall@1 and 7.4% in ROUGE-L.

External Knowledge Acquisition for End-to-End Document-Oriented Dialog Systems

Recent advances in large pre-trained language models (PLMs) lead to impressive gains on natural language understanding (NLU) tasks with task-specific fine-tuning. However, directly fine-tuning PLMs heavily relies on sufficient labeled training instances, which are usually hard to obtain. Prompt-based tuning on PLMs has shown to be powerful for various downstream few-shot tasks. Existing works studying prompt-based tuning for few-shot NLU tasks mainly focus on deriving proper label words with a verbalizer or generating prompt templates to elicit semantics from PLMs. In addition, conventional data augmentation strategies such as synonym substitution are also widely adopted in low-resource scenarios. However, the improvements they bring to prompt-based few-shot learning have been demonstrated to be marginal. Thus, an important research question arises as follows: how to design effective data augmentation methods for prompt-based few-shot tuning? To this end, considering the label semantics are essential in prompt-based tuning, we propose a novel label-guided data augmentation framework PromptDA, which exploits the enriched label semantic information for data augmentation. Extensive experiment results on few-shot text classification tasks show that our proposed framework achieves superior performances by effectively leveraging label semantics and data augmentation for natural language understanding.

PromptDA: Label-guided Data Augmentation for Prompt-based Few Shot Learners

The rise in hateful and offensive language directed at other users is one of the adverse side effects of the increased use of social networking platforms. This could make it difficult for human moderators to review tagged comments filtered by classification systems. To help address this issue, we present the ViHOS (Vietnamese Hate and Offensive Spans) dataset, the first human-annotated corpus containing 26k spans on 11k comments. We also provide definitions of hateful and offensive spans in Vietnamese comments as well as detailed annotation guidelines. Besides, we conduct experiments with various state-of-the-art models. Specifically, XLM-R\_Large achieved the best F1-scores in Single span detection and All spans detection, while PhoBERT\_Large obtained the highest in Multiple spans detection. Finally, our error analysis demonstrates the difficulties in detecting specific types of spans in our data for future research. Our dataset is released on GitHub.

Downloads

Next from EACL 2023

Learning to Ignore Adversarial Attacks