Ireland

Pre-trained models have shown very good performances on a number of question answering benchmarks especially when fine-tuned on multiple question answering datasets at once. In this work, we propose an approach for generating a fine-tuning dataset thanks to a rule-based algorithm that generates questions and answers from unannotated sentences. We show that the state-of-the-art model UnifiedQA can greatly benefit from such a system on a multiple-choice benchmark about physics, biology and chemistry it has never been trained on. We further show that improved performances may be obtained by selecting the most challenging distractors (wrong answers), with a dedicated ranker based on a pretrained RoBERTa model.

ACL 2022

Unsupervised multiple-choice question generation for out-of-domain Q&amp;A fine-tuning

distractor selection

unifiedqa

question answering

# Welcome everyone to ACL 2022!

The 60th Annual Meeting of the Association for Computational Linguistics is taking place May 22-27, 2022 as a hybrid event, in Dublin and online. We are happy to welcome all of you to this anniversary edition with an almost 50-50 in-person and virtual participation. 
The main conference program features oral presentations, in-person and virtual posters and demo sessions, a plenary session for our best paper presentations and awards, three amazing keynote events and two new initiatives of invited talks: Spotlight Talks for Young Rising Stars (STIRS) and The Next Big Idea Talks. Posters (including Findings of ACL 2022) and demos are grouped by areas for both the in-person and the virtual sessions. For the virtual component, the talks will be on Zoom and the posters and the demos will be in GatherTown. The Student Research Workshop will have an oral session and a poster session as part of Poster Session 1. The program also features eight Tutorials and 28 Workshops. 

 
We wish you a wonderful conference! 
[**The ACL 2022 Organizing Committee**](https://www.2022.aclweb.org/organisers)
 
[**Conference Handbook**](https://drive.google.com/file/d/1_BUCMfhMVrjG9E2e71aHdHeE28KSje0l/view?usp=sharing) 
[**Mini Handbook**](https://drive.google.com/file/d/1qlBKl0wzmlVF1oCeMQl3BahLd9nLP5Ce/view?usp=sharing) 
[**Posters and Demo guides**](https://drive.google.com/file/d/1UucMAoCNncIOaH1rMMDa0owuG9qgvJTG/view?usp=sharing)

The Association for Computational Linguistics (ACL) is the premier international scientific and professional society for people working on computational problems involving human language, a field often referred to as either computational linguistics or natural language processing (NLP). 

Unsupervised multiple-choice question generation for out-of-domain Q&A fine-tuning

findings / work in progress

∞-Former: Infinite Memory Transformer

We introduce CaMEL (Case Marker Extraction without Labels), a novel and challenging task in computational morphology that is especially relevant for low-resource languages. We propose a first model for CaMEL that uses a massively multilingual corpus to extract case markers in 83 languages based only on a noun phrase chunker and an alignment system. To evaluate CaMEL, we automatically construct a silver standard from UniMorph. The case markers extracted by our model can be used to detect and visualise similarities and differences between the case systems of different languages as well as to annotate fine-grained deep cases in languages in which they are not overtly marked.

CaMEL: Case Marker Extraction without Labels

Coreference resolution models are typically developed with the OntoNotes dataset. However, these models may not quickly generalize to new, low-resource domains. In our paper, we mitigate this problem with active learning by sampling a small subset of data for annotators to label. While active learning is well-defined for classification tasks, its application to coreference resolution is neither well-defined nor fully understood. Our work explores the type of spans to label according to different sources of model uncertainty. We also investigate whether to label coreference of spans that belong to the same document or across different contexts. In our experiments, we run simulations to compare different active learning strategies and hold a user study to understand the effects of reading on labeling.

Adapting Coreference Resolution Models through Active Learning

Recent work has shown pre-trained language models capture social biases from the large amounts of text they are trained on. This has attracted attention to developing techniques that mitigate such biases. In this work, we perform an empirical survey of five recently proposed bias mitigation techniques: Counterfactual Data Augmentation (CDA), Dropout, Iterative Nullspace Projection, Self-Debias, and SentenceDebias. We quantify the effectiveness of each technique using three intrinsic bias benchmarks while also measuring the impact of these techniques on a model's language modeling ability, as well as its performance on downstream NLU tasks. We experimentally find that: (1) Self-Debias is the strongest debiasing technique, obtaining improved scores on all bias benchmarks;
(2) Current debiasing techniques perform less consistently when mitigating non-gender biases; And (3) improvements on bias benchmarks such as StereoSet and CrowS-Pairs by using debiasing strategies are often accompanied by a decrease in language modeling ability, making it difficult to determine whether the bias mitigation was effective

An Empirical Survey of the Effectiveness of Debiasing Techniques for Pre-trained Language Models

We address the problem of identifying misogyny in tweets in mono and multilingual settings in three languages: English, Italian and Spanish.
We explore model variations considering single and multiple languages both in the pre-training of the transformer and in the training of the downstream task to explore the feasibility of detecting misogyny through a transfer learning approach across multiple languages. That is, we train monolingual transformers with monolingual data and multilingual transformers with both monolingual and multilingual data. Our
models reach state-of-the-art performance on all three languages. The single-language BERT models perform the best, closely followed by different configurations of multilingual BERT models. The performance drops in zero-shot
classification across languages. Our error analysis shows that multilingual and monolingual models tend to make the same mistakes.

 A Checkpoint on Multilingual Misogyny Identification

Automatic speech recognition (ASR) has evolved from a pipeline architecture with pronunciation dictionaries, phonetic features and language models to the end-to-end systems performing a direct translation from a raw waveform into a word sequence. With the increase in accuracy and the availability of pre-trained models, the ASR systems are now omnipresent in our daily applications. On the other hand, the models' interpretability and their computational cost have become more challenging, particularly when dealing with less-common languages or identifying regional variations of speakers%\cite{koenecke2020} no citations in abstracts
. This research proposal will follow a four-stage process: 1) Proving an overview of acoustic features and feature extraction algorithms; 2) Exploring current ASR models, tools, and performance assessment techniques; 3) Aligning features with interpretable phonetic transcripts; and 4) Designing a prototype ARPOCA to increase awareness of regional language variation and improve models feedback by developing a semi-automatic acoustic features extraction using PRAAT in conjunction with phonetic transcription.

 Discourse on ASR Measurement: Introducing the ARPOCA Assessment Tool

We present NESL (the Neuro-Episodic Schema Learner), an event schema learning system that combines large language models, FrameNet parsing, a powerful logical representation of language, and a set of simple behavioral schemas meant to bootstrap the learning process. In lieu of a pre-made corpus of stories, our dataset is a continuous feed of "situation samples" from a pre-trained language model, which are then parsed into FrameNet frames, mapped into simple behavioral schemas, and combined and generalized into complex, hierarchical schemas for a variety of everyday scenarios. We show that careful sampling from the language model can help emphasize stereotypical properties of situations and de-emphasize irrelevant details, and that the resulting schemas specify situations more comprehensively than those learned by other systems.

 Mining Logical Event Schemas From Pre-Trained Language Models

Transformers have achieved state-of-the-art results across multiple NLP tasks. However, the self-attention mechanism complexity scales quadratically with the sequence length, creating an obstacle for tasks involving long sequences, like in the speech domain. In this paper, we discuss the usefulness of self-attention for Direct Speech Translation. First, we analyze the layer-wise token contributions in the self-attention of the encoder, unveiling local diagonal patterns. To prove that some attention weights are avoidable, we propose to substitute the standard self-attention with a local efficient one, setting the amount of context used based on the results of the analysis. With this approach, our model matches the baseline performance, and improves the efficiency by skipping the computation of those weights that standard attention discards.

 On the Locality of Attention in Direct Speech Translation

In this work we put forward to combine pre-
trained knowledge base graph embeddings
with transformer based language models to
improve performance on the sentential Rela-
tion Extraction task in natural language pro-
cessing. Our proposed model is based on a
simple variation of existing models to incor-
porate off-task pre-trained graph embeddings
with an on-task finetuned BERT encoder. We
perform a detailed statistical evaluation of the
model on standard datasets. We provide ev-
idence that the added graph embeddings im-
prove the performance, making such a simple
approach competitive with the state-of-the-art
models that perform explicit on-task training
of the graph embeddings. Furthermore, we ob-
serve for the underlying BERT model an inter-
esting power-law scaling behavior between the
variance of the F1 score obtained for a relation
class and its support in terms of training exam-
ples.

Downloads

Next from ACL 2022

∞-Former: Infinite Memory Transformer

Similar lecture

MuCoT: Multilingual Contrastive Training for Question-Answering in Low-resource Languages