United Arab Emirates

Masked language modeling is widely used for pretraining large language models for natural language understanding (NLU). However, random masking is suboptimal, allocating an equal masking rate for all tokens. In this paper, we propose InforMask, a new unsupervised masking strategy for training masked language models. InforMask exploits Pointwise Mutual Information (PMI) to select the most informative tokens to mask. We further propose two optimizations for InforMask to improve its efficiency. With a one-off preprocessing step, InforMask outperforms random masking and previously proposed masking strategies on the factual recall benchmark LAMA and the question answering benchmark SQuAD v1 and v2.

EMNLP 2022

InforMask: Unsupervised Informative Masking for Language Model Pretraining

informed masking strategy

factual recall

large language models

poster

## Welcome to EMNLP 2022!
I am delighted to welcome you to EMNLP 2022! I believe this conference has been complicated beyond any precedent. Over the past year, it’s been thrilling to see the organization team approach each new puzzle with creativity and enthusiasm. We hope that those participating in Abu Dhabi as well as those joining remotely will leave the conference feeling newly inspired by the program and newly connected to our ever-growing community. Following EMNLP 2021 and major NLP conferences since, EMNLP 2022 is “hybrid,” serving both virtual and in-person participants.

Our key innovations for EMNLP 2022 include:

* EMNLP 2022 is “hybrid” in a second sense, as well: we allowed both direct and rolling review paper submissions, building on the pilot experiment of EMNLP 2021, which considered a small number of ARR submissions. 
* Familiar from NAACL but new to EMNLP, we’ve added an industry track.
* During the conference, “portals” will link virtual poster sessions to in-person conference participants during poster sessions each day.
* The first ACL-family conference in the United Arab Emirates.

 *Message from Noah A. Smith, University of Washington and Allen Institute for AI, Seattle, Washington, USA* 
***EMNLP 2022 General Chair***
 
[![](https://assets.underline.io/uploads/markdown_image/1/image/9eec7d4a287ee18c278b08229290aa83.png)](https://drive.google.com/file/d/1OlPv6QBeo62VVTughj2jkiLeyHd1WnUt/view)
 
[![](https://assets.underline.io/uploads/markdown_image/1/image/a3db7a768409f05192210d98601edb25.png)](https://emnlp2022.rocket.chat/)

To access this site you need to register. Please register [here](https://2022.emnlp.org/registration/).

Register here

Welcome!
EMNLP 2022 will take place in Abu Dhabi from December 7th to December 11th, 2022. And it will be held in hybrid mode, both online and offline.

Despite significant progress in the quality of language generated from abstractive summarization models, these models still exhibit the tendency to hallucinate, i.e., output content not supported by the source document. A number of works have tried to fix---or at least uncover the source of---the problem with limited success. In this paper, we identify a simple criterion under which models are significantly more likely to assign more probability to hallucinated content during generation: high model uncertainty. This finding offers a potential explanation for hallucinations: models default to favoring text with high marginal probability, i.e., high-frequency occurrences in the training set, when uncertain about a continuation. It also motivates possible routes for real-time intervention during decoding to prevent such hallucinations. We propose a decoding strategy that switches to optimizing for pointwise mutual information of the source and target token---rather than purely the probability of the target token---when the model exhibits uncertainty. Experiments on the \xsum dataset show that our method decreases the probability of hallucinated tokens while maintaining the Rouge and BERT-S scores of top-performing decoding strategies.

Mutual Information Alleviates Hallucinations in Abstractive Summarization

Abstractive summarization models typically learn to capture the salient information from scratch implicitly.
Recent literature adds extractive summaries as guidance for abstractive summarization models to provide hints of salient content and achieves better performance.
However, extractive summaries as guidance could be over strict, leading to information loss or noisy signals.
Furthermore, it cannot easily adapt to documents with various abstractiveness.
As the number and allocation of salience content pieces varies, it is hard to find a fixed threshold deciding which content should be included in the guidance.
In this paper, we propose a novel summarization approach with a flexible and reliable salience guidance, namely SEASON (SaliencE Allocation as Guidance for Abstractive SummarizatiON).
SEASON utilizes the allocation of salience expectation to guide abstractive summarization and adapts well to articles in different abstractiveness.
Automatic and human evaluations on two benchmark datasets show that the proposed method is effective and reliable.
Empirical results on more than one million news articles demonstrate a natural fifteen-fifty salience split for news article sentences, providing a useful insight for composing news articles.

Salience Allocation as Guidance for Abstractive Summarization

Question-answering datasets require a broad set of reasoning skills. We show how to use question decompositions to teach language models these broad reasoning skills in a robust fashion. Specifically, we use widely available QDMR representations to programmatically create hard-to-cheat synthetic contexts for real questions in six multi-step reasoning datasets. These contexts are carefully designed to avoid common reasoning shortcuts prevalent in real contexts that prevent models from learning the right skills. This results in a pretraining dataset, named TeaBReaC, containing 525K multi-step questions (with associated formal programs) covering about 900 reasoning patterns. We show that pretraining standard language models (LMs) on TeaBReaC before fine-tuning them on target datasets improves their performance by up to 13 F1 points across 4 multi-step QA datasets, with up to 21 point gain on more complex questions. The resulting models also demonstrate higher robustness, with a 5-8 F1 point improvement on two contrast sets. Furthermore, TeaBReaC pretraining substantially improves model performance and robustness even when starting with numerate LMs pretrained using recent methods (e.g., PReasM, POET). Our work thus shows how to effectively use decomposition-guided contexts to robustly teach multi-step reasoning.

Teaching Broad Reasoning Skills for Multi-Step QA by Generating Hard Contexts

Real-world politically-opinionated memes often rely on figurative language to cloak propaganda and radical ideas to help them spread. It is not only a scientific challenge to develop machine learning models to recognize them in memes, but also sociologically beneficial to understand hidden meanings at scale and raise awareness. These memes are fast-evolving (in both topics and visuals) and it remains unclear whether current multimodal machine learning models are robust to such distribution shifts. To enable future research into this area, we first present FigMemes, a dataset for figurative language classification in politically-opinionated memes. We evaluate the performance of state-of-the-art unimodal and multimodal models and provide comprehensive benchmark results. The key contributions of this proposed dataset include annotations of six commonly used types of figurative language in politically-opinionated memes, and a wide range of topics and visual styles.
We also provide analyses on the ability of multimodal models to generalize across distribution shifts in memes. Our dataset poses unique machine learning challenges and our results show that current models have significant room for improvement in both performance and robustness to distribution shifts.

FigMemes: A Dataset for Figurative Language Identification in Politically-Opinionated Memes

Radiology report generation systems have the potential to reduce the workload of radiologists by automatically describing the findings in medical images.
To broaden the application of the report generation system, the system should generate reports that are not only factually accurate but also chronologically consistent, describing images that are presented in time order, that is, the correct order.
We employ a planning-based radiology report generation system that generates the overall structure of reports as "plans'" prior to generating reports that are accurate and consistent in order.
Additionally, we propose a novel reinforcement learning and inference method, Coordinated Planning (CoPlan), that includes a content planner and a text generator to train and infer in a coordinated manner to alleviate the cascading of errors that are often inherent in planning-based models.
We conducted experiments with single-phase diagnostic reports in which the factual accuracy is critical and multi-phase diagnostic reports in which the description order is critical.
Our proposed CoPlan improves the content order score by 5.1 pt in time series critical scenarios and the clinical factual accuracy F-score by 9.1 pt in time series irrelevant scenarios, compared those of the baseline models without CoPlan.

Factual Accuracy is not Enough: Planning Consistent Description Order for Radiology Report Generation

Pre-trained multilingual models, such as mBERT, XLM-R and mT5, are used to improve the performance on various tasks in low-resource languages via cross-lingual transfer. In this framework, English is usually seen as the most natural choice for a transfer language (for fine-tuning or continued training of a multilingual pre-trained model), but it has been revealed recently that this is often not the best choice. The success of cross-lingual transfer seems to depend on some properties of languages, which are currently hard to explain. Successful transfer often happens between unrelated languages and it often cannot be explained by data-dependent factors.

In this study, we show that languages written in non-Latin and non-alphabetic scripts (mostly Asian languages) are the best choices for improving performance on the task of Masked Language Modelling (MLM) in a diverse set of 30 low-resource languages and that the success of the transfer is well predicted by our novel measure of Subword Evenness (SuE). Transferring language models over the languages that score low on our measure results in the lowest average perplexity over target low-resource languages. Our correlation coefficients obtained with three different pre-trained multilingual models are consistently higher than all the other predictors, including text-based measures (type-token ratio, entropy) and linguistically motivated choice (genealogical and typological proximity).

Subword Evenness (SuE) as a Predictor of Cross-lingual Transfer to Low-resource Languages

Masked language models like BERT can perform text classification in a zero-shot fashion by reformulating downstream tasks as text infilling. However, this approach is highly sensitive to the template used to prompt the model, yet practitioners are blind when designing them in strict zero-shot settings. In this paper, we propose an alternative mining-based approach for zero-shot learning. Instead of prompting language models, we use regular expressions to mine labeled examples from unlabeled corpora, which can optionally be filtered through prompting, and used to finetune a pretrained model. Our method is more flexible and interpretable than prompting, and outperforms it on a wide range of tasks when using comparable templates. Our results suggest that the success of prompting can partly be explained by the model being exposed to similar examples during pretraining, which can be directly retrieved through regular expressions.

Don't Prompt, Search! Mining-based Zero-Shot Learning with Language Models

Transformer-based pre-trained language models are vocabulary-dependent, mapping by default each token to its corresponding embedding. This one-to-one mapping results into embedding matrices that occupy a lot of memory (i.e. millions of parameters) and grow linearly with the size of the vocabulary. Previous work on on-device transformers dynamically generate token embeddings on-the-fly without embedding matrices using locality-sensitive hashing over morphological information. These embeddings are subsequently fed into transformer layers for text classification. However, these methods are not pre-trained. Inspired by this line of work, we propose HashFormers, a new family of vocabulary-independent pre-trained transformers that support an unlimited vocabulary (i.e. all possible tokens in a corpus) given a substantially smaller fixed-sized embedding matrix. We achieve this by first introducing computationally cheap hashing functions that bucket together individual tokens to embeddings. We also propose three variants that do not require an embedding matrix at all, further reducing the memory requirements. We empirically demonstrate that HashFormers are more memory efficient compared to standard pre-trained transformers while achieving comparable predictive performance when fine-tuned on multiple text classification tasks. For example, our most efficient HashFormer variant has a negligible performance degradation (0.4% on GLUE) using only 99.1K parameters for representing the embeddings compared to 12.3-38M parameters of state-of-the-art models.

HashFormers: Towards Vocabulary-independent Pre-trained Transformers

A growing body of work studies how to answer a question or verify a claim by generating a natural language "proof:” a chain of deductive inferences yielding the answer based on a set of premises. However, these methods can only make sound deductions when they follow from evidence that is given. We propose a new system that can handle the underspecified setting where not all premises are stated at the outset; that is, additional assumptions need to be materialized to prove a claim. By using a natural language generation model to abductively infer a premise given another premise and a conclusion, we can impute missing pieces of evidence needed for the conclusion to be true. Our system searches over two fringes in a bidirectional fashion, interleaving deductive (forward-chaining) and abductive (backward-chaining) generation steps. We sample multiple possible outputs for each step to achieve coverage of the search space, at the same time ensuring correctness by filtering low-quality generations with a round-trip validation procedure. Results on a modified version of the EntailmentBank dataset and a new dataset called Everyday Norms: Why Not? Show that abductive generation with validation can recover premises across in- and out-of-domain settings.

Natural Language Deduction with Incomplete Information

The wide adoption and application of Masked language models~(MLMs) on sensitive data (from legal to medical) necessitates a thorough quantitative investigation into their privacy vulnerabilities. 
Prior attempts at measuring leakage of MLMs via membership inference attacks have been inconclusive, implying potential robustness of MLMs to privacy attacks.
In this work, we posit that prior attempts were inconclusive because they based their attack solely on the MLM's model score. We devise a stronger membership inference attack based on likelihood ratio hypothesis testing that involves an additional reference MLM to more accurately quantify the privacy risks of memorization in MLMs. We show that masked language models are indeed susceptible to likelihood ratio membership inference attacks: Our empirical results, on models trained on medical notes, show that our attack improves the AUC of prior membership inference attacks from $0.66$ to an alarmingly high $0.90$ level.

InforMask: Unsupervised Informative Masking for Language Model Pretraining

Downloads

Next from EMNLP 2022

Mutual Information Alleviates Hallucinations in Abstractive Summarization

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

InforMask: Unsupervised Informative Masking for Language Model Pretraining

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from EMNLP 2022

Mutual Information Alleviates Hallucinations in Abstractive Summarization

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Downloads