We present Mr. TyDi, a multi-lingual benchmark dataset for mono-lingual retrieval in eleven typologically diverse languages, designed to evaluate ranking with learned dense representations. The goal of this resource is to spur research in dense retri
eval techniques in non-English languages, motivated by recent observations that existing techniques for representation learning perform poorly when applied to out-of-distribution data. As a starting point, we provide zero-shot baselines for this new dataset based on a multi-lingual adaptation of DPR that we call mDPR''. Experiments show that although the effectiveness of mDPR is much lower than BM25, dense representations nevertheless appear to provide valuable relevance signals, improving BM25 results in sparse--dense hybrids. In addition to analyses of our results, we also discuss future challenges and present a research agenda in multi-lingual dense retrieval. Mr. TyDi can be downloaded at https://github.com/castorini/mr.tydi.
Punctuation restoration is a fundamental requirement for the readability of text derived from Automatic Speech Recognition (ASR) systems. Most contemporary solutions are limited to predicting only a few of the most frequently occurring marks, such as
periods, commas, and question marks - and only one per word. However, in written language, we deal with a much larger number of punctuation characters (such as parentheses, hyphens, etc.), and their combinations (like parenthesis followed by a dot). Such comprehensive punctuation cannot always be unambiguously reduced to a basic set of the most frequently occurring marks. In this work, we evaluate several methods in the comprehensive punctuation reconstruction task. We conduct experiments on parallel corpora of two different languages, English and Polish - languages with a relatively simple and complex morphology, respectively. We also investigate the influence of building a model on comprehensive punctuation on the quality of the basic punctuation restoration task
Existing techniques for mitigating dataset bias often leverage a biased model to identify biased instances. The role of these biased instances is then reduced during the training of the main model to enhance its robustness to out-of-distribution data
. A common core assumption of these techniques is that the main model handles biased instances similarly to the biased model, in that it will resort to biases whenever available. In this paper, we show that this assumption does not hold in general. We carry out a critical investigation on two well-known datasets in the domain, MNLI and FEVER, along with two biased instance detection methods, partial-input and limited-capacity models. Our experiments show that in around a third to a half of instances, the biased model is unable to predict the main model's behavior, highlighted by the significantly different parts of the input on which they base their decisions. Based on a manual validation, we also show that this estimate is highly in line with human interpretation. Our findings suggest that down-weighting of instances detected by bias detection methods, which is a widely-practiced procedure, is an unnecessary waste of training data. We release our code to facilitate reproducibility and future research.
Generating texts in scientific papers requires not only capturing the content contained within the given input but also frequently acquiring the external information called context. We push forward the scientific text generation by proposing a new ta
sk, namely context-aware text generation in the scientific domain, aiming at exploiting the contributions of context in generated texts. To this end, we present a novel challenging large-scale Scientific Paper Dataset for ConteXt-Aware Text Generation (SciXGen), consisting of well-annotated 205,304 papers with full references to widely-used objects (e.g., tables, figures, algorithms) in a paper. We comprehensively benchmark, using state-of-the-arts, the efficacy of our newly constructed SciXGen dataset in generating description and paragraph. Our dataset and benchmarks will be made publicly available to hopefully facilitate the scientific text generation research.
Abstractive summarization quality had large improvements since recent language pretraining techniques. However, currently there is a lack of datasets for the growing needs of conversation summarization applications. Thus we collected ForumSum, a dive
rse and high-quality conversation summarization dataset with human written summaries. The conversations in ForumSum dataset are collected from a wide variety of internet forums. To make the dataset easily expandable, we also release the process of dataset creation. Our experiments show that models trained on ForumSum have better zero-shot and few-shot transferability to other datasets than the existing large chat summarization dataset SAMSum. We also show that using a conversational corpus for pre-training improves the quality of the chat summarization model.
As the world continues to fight the COVID-19 pandemic, it is simultaneously fighting an infodemic' -- a flood of disinformation and spread of conspiracy theories leading to health threats and the division of society. To combat this infodemic, there i
s an urgent need for benchmark datasets that can help researchers develop and evaluate models geared towards automatic detection of disinformation. While there are increasing efforts to create adequate, open-source benchmark datasets for English, comparable resources are virtually unavailable for German, leaving research for the German language lagging significantly behind. In this paper, we introduce the new benchmark dataset FANG-COVID consisting of 28,056 real and 13,186 fake German news articles related to the COVID-19 pandemic as well as data on their propagation on Twitter. Furthermore, we propose an explainable textual- and social context-based model for fake news detection, compare its performance to black-box'' models and perform feature ablation to assess the relative importance of human-interpretable features in distinguishing fake news from authentic news.
Many datasets have been created for training reading comprehension models, and a natural question is whether we can combine them to build models that (1) perform better on all of the training datasets and (2) generalize and transfer better to new dat
asets. Prior work has addressed this goal by training one network simultaneously on multiple datasets, which works well on average but is prone to over- or under-fitting different sub- distributions and might transfer worse compared to source models with more overlap with the target dataset. Our approach is to model multi-dataset question answering with an ensemble of single-dataset experts, by training a collection of lightweight, dataset-specific adapter modules (Houlsby et al., 2019) that share an underlying Transformer model. We find that these Multi-Adapter Dataset Experts (MADE) outperform all our baselines in terms of in-distribution accuracy, and simple methods based on parameter-averaging lead to better zero-shot generalization and few-shot transfer performance, offering a strong and versatile starting point for building new reading comprehension systems.
Much of recent progress in NLU was shown to be due to models' learning dataset-specific heuristics. We conduct a case study of generalization in NLI (from MNLI to the adversarially constructed HANS dataset) in a range of BERT-based architectures (ada
pters, Siamese Transformers, HEX debiasing), as well as with subsampling the data and increasing the model size. We report 2 successful and 3 unsuccessful strategies, all providing insights into how Transformer-based models learn to generalize.
In this paper, we present the first multilingual FAQ dataset publicly available. We collected around 6M FAQ pairs from the web, in 21 different languages. Although this is significantly larger than existing FAQ retrieval datasets, it comes with its o
wn challenges: duplication of content and uneven distribution of topics. We adopt a similar setup as Dense Passage Retrieval (DPR) and test various bi-encoders on this dataset. Our experiments reveal that a multilingual model based on XLM-RoBERTa achieves the best results, except for English. Lower resources languages seem to learn from one another as a multilingual model achieves a higher MRR than language-specific ones. Our qualitative analysis reveals the brittleness of the model on simple word changes. We publicly release our dataset, model, and training script.
Bidirectional Encoder Representations from Transformers (BERT) has achieved state-of-the-art performances on several text classification tasks, such as GLUE and sentiment analysis. Recent work in the legal domain started to use BERT on tasks, such as
legal judgement prediction and violation prediction. A common practise in using BERT is to fine-tune a pre-trained model on a target task and truncate the input texts to the size of the BERT input (e.g. at most 512 tokens). However, due to the unique characteristics of legal documents, it is not clear how to effectively adapt BERT in the legal domain. In this work, we investigate how to deal with long documents, and how is the importance of pre-training on documents from the same domain as the target task. We conduct experiments on the two recent datasets: ECHR Violation Dataset and the Overruling Task Dataset, which are multi-label and binary classification tasks, respectively. Importantly, on average the number of tokens in a document from the ECHR Violation Dataset is more than 1,600. While the documents in the Overruling Task Dataset are shorter (the maximum number of tokens is 204). We thoroughly compare several techniques for adapting BERT on long documents and compare different models pre-trained on the legal and other domains. Our experimental results show that we need to explicitly adapt BERT to handle long documents, as the truncation leads to less effective performance. We also found that pre-training on the documents that are similar to the target task would result in more effective performance on several scenario.
