Pre-trained language models like BERT and its variants have recently achieved impressive performance in various natural language understanding tasks. However, BERT heavily relies on the global self-attention block and thus suffers large memory footprint and computation cost. Although all its attention heads query on the whole input sequence for generating the attention map from a global perspective, we observe some heads only need to learn local dependencies, which means the existence of computation redundancy. We therefore propose a novel span-based dynamic convolution to replace these self-attention heads to directly model local dependencies. The novel convolution heads, together with the rest self-attention heads, form a new mixed attention block that is more efficient at both global and local context learning. We equip BERT with this mixed attention design and build a ConvBERT model. Experiments have shown that ConvBERT significantly outperforms BERT and its variants in various downstream tasks, with lower training cost and fewer model parameters. Remarkably, ConvBERTbase model achieves 86.4 GLUE score, 0.7 higher than ELECTRAbase, while using less than 1/4 training cost. Code and pre-trained models will be released.
National Library of Norway（挪威国家图书馆）；Telefónica I+D（西班牙电信研发公司）；Foqum（Foqum公司）；Narrativa（Narrativa公司）
The pre-training of large language models usually requires massive amounts of resources, both in terms of computation and data. Frequently used web sources such as Common Crawl might contain enough noise to make this pre-training sub-optimal. In this work, we experiment with different sampling methods from the Spanish version of mC4, and present a novel data-centric technique which we name $\textit{perplexity sampling}$ that enables the pre-training of language models in roughly half the amount of steps and using one fifth of the data. The resulting models are comparable to the current state-o…
Pre-trained Transformer-based neural language models, such as BERT, have achieved remarkable results on varieties of NLP tasks. Recent works have shown that attention-based models can benefit from more focused attention over local regions. Most of them restrict the attention scope within a linear span, or confine to certain tasks such as machine translation and question answering. In this paper, we propose a syntax-aware local attention, where the attention scopes are restrained based on the distances in the syntactic structure. The proposed syntax-aware local attention can be integrated with …
Université Paris-Saclay（巴黎-萨克雷大学）；CNRS, LISN（法国国家科学研究中心，LISN）；University of Helsinki（赫尔辛基大学）
Pre-trained language models (LMs) encode rich information about linguistic structure but their knowledge about lexical polysemy remains unclear. We propose a novel experimental setup for analysing this knowledge in LMs specifically trained for different languages (English, French, Spanish and Greek) and in multilingual BERT. We perform our analysis on datasets carefully designed to reflect different sense distributions, and control for parameters that are highly correlated with polysemy such as frequency and grammatical category. We demonstrate that BERT-derived representations reflect words' …
This paper describes our system for SemEval-2021 Task 5 on Toxic Spans Detection. We developed ensemble models using BERT-based neural architectures and post-processing to combine tokens into spans. We evaluated several pre-trained language models using various ensemble techniques for toxic span identification and achieved sizable improvements over our baseline fine-tuned BERT models. Finally, our system obtained a F1-score of 67.55% on test data.
Exploring Large Language Models for Hate Speech Detection in Rioplatense Spanish
Instituto de Cs. de la Computación（计算机科学研究所）；CONICET（阿根廷科学技术研究委员会）；UBA（布宜诺斯艾利斯大学）；Instituto Gino Germani（吉诺·格尔马研究所）；FSOC（社会科学学院）；Departamento de Computación, FCEN UBA（布宜诺斯艾利斯大学精确与自然科学学院计算机系）
Our goal of patent claim generation is to realize "augmented inventing" for inventors by leveraging latest Deep Learning techniques. We envision the possibility of building an "auto-complete" function for inventors to conceive better inventions in the era of artificial intelligence. In order to generate patent claims with good quality, a fundamental question is how to measure it. We tackle the problem from a perspective of claim span relevancy. Patent claim language was rarely explored in the NLP field. It is unique in its own way and contains rich explicit and implicit human annotations. In t…
University of Information Technology（信息技术大学）；Vietnam National University, Ho Chi Minh City（胡志明市越南国家大学）
In this paper, we propose a span labeling approach to model n-gram information for Vietnamese word segmentation, namely SPAN SEG. We compare the span labeling approach with the conditional random field by using encoders with the same architecture. Since Vietnamese and Chinese have similar linguistic phenomena, we evaluated the proposed method on the Vietnamese treebank benchmark dataset and five Chinese benchmark datasets. Through our experimental results, the proposed approach SpanSeg achieves higher performance than the sequence tagging approach with the state-of-the-art F-score of 98.31% on…
International Digital Economy Academy（国际数字经济学院）；South China University of Technology（华南理工大学）
Even as pre-trained language models share a semantic encoder, natural language understanding suffers from a diversity of output schemas. In this paper, we propose UBERT, a unified bidirectional language understanding model based on BERT framework, which can universally model the training objects of different NLU tasks through a biaffine network. Specifically, UBERT encodes prior knowledge from various aspects, uniformly constructing learning representations across multiple NLU tasks, which is conducive to enhancing the ability to capture common semantic understanding. By using the biaffine to …
College of Arts and Sciences, School of Information Studies, Syracuse University（雪城大学艺术与科学学院、信息研究学院）
This paper describes the BERT-based models proposed for two subtasks in SemEval-2020 Task 11: Detection of Propaganda Techniques in News Articles. We first build the model for Span Identification (SI) based on SpanBERT, and facilitate the detection by a deeper model and a sentence-level representation. We then develop a hybrid model for the Technique Classification (TC). The hybrid model is composed of three submodels including two BERT models with different training methods, and a feature-based Logistic Regression model. We endeavor to deal with imbalanced dataset by adjusting cost function. …
University of Washington（华盛顿大学）；George Mason University（乔治梅森大学）
Medical imaging is critical to the diagnosis and treatment of numerous medical problems, including many forms of cancer. Medical imaging reports distill the findings and observations of radiologists, creating an unstructured textual representation of unstructured medical images. Large-scale use of this text-encoded information requires converting the unstructured text to a structured, semantic representation. We explore the extraction and normalization of anatomical information in radiology reports that is associated with radiological findings. We investigate this extraction and normalization …
University of Chile（智利大学）；National Center for Artificial Intelligence (CENIA)（国家人工智能中心（CENIA））；Millennium Institute for Foundational Research on Data (IMFD)（千禧数据基础研究所（IMFD））；KU Leuven（鲁汶大学）
In recent years there have been considerable advances in pre-trained language models, where non-English language versions have also been made available. Due to their increasing use, many lightweight versions of these models (with reduced parameters) have also been released to speed up training and inference times. However, versions of these lighter models (e.g., ALBERT, DistilBERT) for languages other than English are still scarce. In this paper we present ALBETO and DistilBETO, which are versions of ALBERT and DistilBERT pre-trained exclusively on Spanish corpora. We train several versions of…
With the ever-increasing availability of digital information, toxic content is also on the rise. Therefore, the detection of this type of language is of paramount importance. We tackle this problem utilizing a combination of a state-of-the-art pre-trained language model (CharacterBERT) and a traditional bag-of-words technique. Since the content is full of toxic words that have not been written according to their dictionary spelling, attendance to individual characters is crucial. Therefore, we use CharacterBERT to extract features based on the word characters. It consists of a CharacterCNN mod…
NTT Communication Science Laboratories, NTT Corporation（日本电信电话株式会社NTT通信科学实验室）
We present a novel supervised word alignment method based on cross-language span prediction. We first formalize a word alignment problem as a collection of independent predictions from a token in the source sentence to a span in the target sentence. As this is equivalent to a SQuAD v2.0 style question answering task, we then solve this problem by using multilingual BERT, which is fine-tuned on a manually created gold word alignment data. We greatly improved the word alignment accuracy by adding the context of the token to the question. In the experiments using five word alignment datasets amon…
Institute of Information Engineering, Chinese Academy of Sciences（中国科学院信息工程研究所）；School of Cyber Security, University of Chinese Academy of Sciences（中国科学院大学网络空间安全学院）
Recently, generative methods have been widely used in keyphrase prediction, thanks to their capability to produce both present keyphrases that appear in the source text and absent keyphrases that do not match any source text. However, the absent keyphrases are generated at the cost of the performance on present keyphrase prediction, since previous works mainly use generative models that rely on the copying mechanism and select words step by step. Besides, the extractive model that directly extracts a text span is more suitable for predicting the present keyphrase. Considering the different cha…
Universidad de Buenos Aires（布宜诺斯艾利斯大学）；CONICET（国家科学技术研究委员会）
Since BERT appeared, Transformer language models and transfer learning have become state-of-the-art for Natural Language Understanding tasks. Recently, some works geared towards pre-training specially-crafted models for particular domains, such as scientific papers, medical documents, user-generated texts, among others. These domain-specific models have been shown to improve performance significantly in most tasks. However, for languages other than English such models are not widely available. In this work, we present RoBERTuito, a pre-trained language model for user-generated text in Spanish,…
Toxicity detection of text has been a popular NLP task in the recent years. In SemEval-2021 Task-5 Toxic Spans Detection, the focus is on detecting toxic spans within passages. Most state-of-the-art span detection approaches employ various techniques, each of which can be broadly classified into Token Classification or Span Prediction approaches. In our paper, we explore simple versions of both of these approaches and their performance on the task. Specifically, we use BERT-based models -- BERT, RoBERTa, and SpanBERT for both approaches. We also combine these approaches and modify them to brin…
Beijing Academy of Artificial Intelligence（北京智源人工智能研究院）；Tsinghua University（清华大学）；Shanghai Qi Zhi Institute（上海期智研究院）；MIT CSAIL（麻省理工学院计算机科学与人工智能实验室）
There have been various types of pretraining architectures including autoencoding models (e.g., BERT), autoregressive models (e.g., GPT), and encoder-decoder models (e.g., T5). However, none of the pretraining frameworks performs the best for all tasks of three main categories including natural language understanding (NLU), unconditional generation, and conditional generation. We propose a General Language Model (GLM) based on autoregressive blank infilling to address this challenge. GLM improves blank filling pretraining by adding 2D positional encodings and allowing an arbitrary order to pre…
In this paper, we propose a method to extract bilingual texts automatically from noisy parallel corpora by framing the problem as a token-level span prediction, such as SQuAD-style Reading Comprehension. To extract a span of the target document that is a translation of a given source sentence (span), we use either QANet or multilingual BERT. QANet can be trained for a specific parallel corpus from scratch, while multilingual BERT can utilize pre-trained multilingual representations. For the span prediction method using QANet, we introduce a total optimization method using integer linear progra…
We present a highly parameter efficient approach for Question Answering that significantly reduces the need for extended BERT fine-tuning. Our method uses information from the hidden state activations of each BERT transformer layer, which is discarded during typical BERT inference. Our best model achieves maximal BERT performance at a fraction of the training time and GPU or TPU expense. Performance is further improved by ensembling our model with BERTs predictions. Furthermore, we find that near optimal performance can be achieved for QA span annotation using less training data. Our experimen…
Facebook AI Research（Facebook人工智能研究院）；University of Washington（华盛顿大学）；Princeton University（普林斯顿大学）；Allen Institute of Artificial Intelligence（艾伦人工智能研究所）
We present SpanBERT, a pre-training method that is designed to better represent and predict spans of text. Our approach extends BERT by (1) masking contiguous random spans, rather than random tokens, and (2) training the span boundary representations to predict the entire content of the masked span, without relying on the individual token representations within it. SpanBERT consistently outperforms BERT and our better-tuned baselines, with substantial gains on span selection tasks such as question answering and coreference resolution. In particular, with the same training data and model size a…
International Digital Economy Academy (IDEA)（国际数字经济研究院（IDEA））；Beijing University of Posts and Telecommunications（北京邮电大学）；Nanjing University（南京大学）
Bidirectional Encoder Representations from Transformers or BERT~\cite{devlin-etal-2019-bert} has been one of the base models for various NLP tasks due to its remarkable performance. Variants customized for different languages and tasks are proposed to further improve the performance. In this work, we investigate supervised continued pre-training~\cite{gururangan-etal-2020-dont} on BERT for Chinese topic classification task. Specifically, we incorporate prompt-based learning and contrastive learning into the pre-training. To adapt to the task of Chinese topic classification, we collect around 2…
BERT (Bidirectional Encoder Representations from Transformers) and related pre-trained Transformers have provided large gains across many language understanding tasks, achieving a new state-of-the-art (SOTA). BERT is pre-trained on two auxiliary tasks: Masked Language Model and Next Sentence Prediction. In this paper we introduce a new pre-training task inspired by reading comprehension to better align the pre-training from memorization to understanding. Span Selection Pre-Training (SSPT) poses cloze-like training instances, but rather than draw the answer from the model's parameters, it is se…
Since the appearance of BERT, recent works including XLNet and RoBERTa utilize sentence embedding models pre-trained by large corpora and a large number of parameters. Because such models have large hardware and a huge amount of data, they take a long time to pre-train. Therefore it is important to attempt to make smaller models that perform comparatively. In this paper, we trained a Korean-specific model KR-BERT, utilizing a smaller vocabulary and dataset. Since Korean is one of the morphologically rich languages with poor resources using non-Latin alphabets, it is also important to capture l…
Northeastern University（东北大学）；Google Research（谷歌研究院）；University of Pennsylvania（宾夕法尼亚大学）
While composing a new document, anything from a news article to an email or essay, authors often utilize direct quotes from a variety of sources. Although an author may know what point they would like to make, selecting an appropriate quote for the specific context may be time-consuming and difficult. We therefore propose a novel context-aware quote recommendation system which utilizes the content an author has already written to generate a ranked list of quotable paragraphs and spans of tokens from a given source document. We approach quote recommendation as a variant of open-domain question …
East China Normal University（华东师范大学）；Fudan University（复旦大学）；Pingan Health Tech（平安健康医疗科技）；Ping An Health Cloud Company Limited（平安健康云有限公司）；Ping An International Smart City Technology Co., Ltd（平安国际智慧城市科技股份有限公司）
Although BERT based relation classification (RC) models have achieved significant improvements over the traditional deep learning models, it seems that no consensus can be reached on what is the optimal architecture. Firstly, there are multiple alternatives for entity span identification. Second, there are a collection of pooling operations to aggregate the representations of entities and contexts into fixed length vectors. Third, it is difficult to manually decide which feature vectors, including their interactions, are beneficial for classifying the relation types. In this work, we design a …
Even as pre-trained language encoders such as BERT are shared across many tasks, the output layers of question answering, text classification, and regression models are significantly different. Span decoders are frequently used for question answering, fixed-class, classification layers for text classification, and similarity-scoring layers for regression tasks, We show that this distinction is not necessary and that all three can be unified as span extraction. A unified, span-extraction approach leads to superior or comparable performance in supplementary supervised pre-trained, low-data, and …
RheinMain University of Applied Sciences（莱茵美因应用科学大学）
We introduce SpERT, an attention model for span-based joint entity and relation extraction. Our key contribution is a light-weight reasoning on BERT embeddings, which features entity recognition and filtering, as well as relation classification with a localized, marker-free context representation. The model is trained using strong within-sentence negative samples, which are efficiently extracted in a single BERT pass. These aspects facilitate a search over all spans in the sentence. In ablation studies, we demonstrate the benefits of pre-training, strong negative sampling and localized context…
New York University（纽约大学）；Verisk Analytics（Verisk分析公司）
We investigate the extent to which individual attention heads in pretrained transformer language models, such as BERT and RoBERTa, implicitly capture syntactic dependency relations. We employ two methods---taking the maximum attention weight and computing the maximum spanning tree---to extract implicit dependency relations from the attention weights of each layer/head, and compare them to the ground-truth Universal Dependency (UD) trees. We show that, for some UD relation types, there exist heads that can recover the dependency type significantly better than baselines on parsed English text, s…
In this paper, we propose Multi$^2$OIE, which performs open information extraction (open IE) by combining BERT with multi-head attention. Our model is a sequence-labeling system with an efficient and effective argument extraction method. We use a query, key, and value setting inspired by the Multimodal Transformer to replace the previously used bidirectional long short-term memory architecture with multi-head attention. Multi$^2$OIE outperforms existing sequence-labeling systems with high computational efficiency on two benchmark evaluation datasets, Re-OIE2016 and CaRB. Additionally, we apply…
The pre-trained BERT model achieves a remarkable state of the art across a wide range of tasks in natural language processing. For solving the gender bias in gendered pronoun resolution task, I propose a novel neural network model based on the pre-trained BERT. This model is a type of mention score classifier and uses an attention mechanism with no parameters to compute the contextual representation of entity span, and a vector to represent the triple-wise semantic similarity among the pronoun and the entities. In stage 1 of the gendered pronoun resolution task, a variant of this model, traine…
We introduce a new scientific named entity recognizer called SEPT, which stands for Span Extractor with Pre-trained Transformers. In recent papers, span extractors have been demonstrated to be a powerful model compared with sequence labeling models. However, we discover that with the development of pre-trained language models, the performance of span extractors appears to become similar to sequence labeling models. To keep the advantages of span representation, we modified the model by under-sampling to balance the positive and negative samples and reduce the search space. Furthermore, we simp…
Indian Institute of Technology Kanpur (IITK)（印度理工学院坎普尔分校）
This paper describes our submissions to SemEval 2020 Task 11: Detection of Propaganda Techniques in News Articles for each of the two subtasks of Span Identification and Technique Classification. We make use of pre-trained BERT language model enhanced with tagging techniques developed for the task of Named Entity Recognition (NER), to develop a system for identifying propaganda spans in the text. For the second subtask, we incorporate contextual features in a pre-trained RoBERTa model for the classification of propaganda techniques. We were ranked 5th in the propaganda technique classification…
Nanyang Technological University（南洋理工大学）；Salesforce Research Asia（Salesforce亚洲研究院）；Institute for Infocomm Research, A-STAR（A*STAR资讯通信研究院）
We propose a novel constituency parsing model that casts the parsing problem into a series of pointing tasks. Specifically, our model estimates the likelihood of a span being a legitimate tree constituent via the pointing score corresponding to the boundary words of the span. Our parsing model supports efficient top-down decoding and our learning objective is able to enforce structural consistency without resorting to the expensive CKY inference. The experiments on the standard English Penn Treebank parsing task show that our method achieves 92.78 F1 without using pre-trained models, which is …
University of California, San Diego（加州大学圣地亚哥分校）；IBM Research（IBM研究院）
Knowledge-enhanced pre-trained models for language representation have been shown to be more effective in knowledge base construction tasks (i.e.,~relation extraction) than language models such as BERT. These knowledge-enhanced language models incorporate knowledge into pre-training to generate representations of entities or relationships. However, existing methods typically represent each entity with a separate embedding. As a result, these methods struggle to represent out-of-vocabulary entities and a large amount of parameters, on top of their underlying token models (i.e.,~the transformer)…
