The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called “ImageNet”, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond.

/pdf/imagenet-a-large-scale-hierarchical-image-database-1e8f663s1h.pdf

ImageNet: A large-scale hierarchical image database

We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models, BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. 
BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5% (7.7% point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement).

BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

Language model pretraining has led to significant performance gains but careful comparison between different approaches is challenging. Training is computationally expensive, often done on private datasets of different sizes, and, as we will show, hyperparameter choices have significant impact on the final results. We present a replication study of BERT pretraining (Devlin et al., 2019) that carefully measures the impact of many key hyperparameters and training data size. We find that BERT was significantly undertrained, and can match or exceed the performance of every model published after it. Our best model achieves state-of-the-art results on GLUE, RACE and SQuAD. These results highlight the importance of previously overlooked design choices, and raise questions about the source of recently reported improvements. We release our models and code.

/pdf/roberta-a-robustly-optimized-bert-pretraining-approach-2rj0bdyhim.pdf

RoBERTa: A Robustly Optimized BERT Pretraining Approach

Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general.

/pdf/language-models-are-few-shot-learners-2fg8gvia7m.pdf

Language Models are Few-Shot Learners

How to Do Things With Words

This paper describes the PASCAL Network of Excellence first Recognising Textual Entailment (RTE-1) Challenge benchmark 1 . The RTE task is defined as recognizing, given two text fragments, whether the meaning of one text can be inferred (entailed) from the other. This application-independent task is suggested as capturing major inferences about the variability of semantic expression which are commonly needed across multiple applications. The Challenge has raised noticeable attention in the research community, attracting 17 submissions from diverse groups, suggesting the generic relevance of the task.

The PASCAL Recognising Textual Entailment Challenge

This paper presents the Third PASCAL Recognising Textual Entailment Challenge (RTE-3), providing an overview of the dataset creating methodology and the submitted systems. In creating this year's dataset, a number of longer texts were introduced to make the challenge more oriented to realistic scenarios. Additionally, a pool of resources was offered so that the participants could share common tools. A pilot task was also set up, aimed at differentiating unknown entailments from identified contradictions and providing justifications for overall system decisions. 26 participants submitted 44 runs, using different approaches and generally presenting new entailment models and achieving higher scores than in the previous challenges.

/pdf/the-third-pascal-recognizing-textual-entailment-challenge-1m9o1mxy76.pdf

The Third PASCAL Recognizing Textual Entailment Challenge

This volume brings together ontology learning, knowledge acquisition and other related topics It presents current research in ontology learning, addressing three perspectives The first perspective looks at methodologies that have been proposed to automatically extract information from texts and to give a structured organization to such knowledge, including approaches based on machine learning techniques Then there are evaluation methods for ontology learning, aiming at defining procedures and metrics for a quantitative evaluation of the ontology learning task; and finally application scenarios that make ontology learning a challenging area in the context of real applications such as bio-informatics According to the three perspectives mentioned above, the book is divided into three sections, each including a selection of papers addressing respectively the methods, the applications and the evaluation of ontology learning approaches

Ontology Learning from Text: Methods, Evaluation and Applications

In this paper, we present a lexical resource where WordNet synsets are annotated with Subject Field Codes. We discuss both the methodological issues we dealt with and the annotation techniques used. A quantitative analysis of the resource coverage, as well as a qualitative evaluation of the proposed annotations, are reported.

/pdf/integrating-subject-field-codes-into-wordnet-13ur9lxc22.pdf

Integrating Subject Field Codes into WordNet

This volume brings together a collection of extended versions of selected papers from two workshops on ontology learning, knowledge acquisition and related topics that were organized in the context of the European Conference on Artificial Intelligence (ECAI) 2004 and the International Conference on Knowledge Engineering and Management (EKAW) 2004. The volume presents current research in ontology learning, addressing three perspectives: methodologies that have been proposed to automatically extract information from texts and to give a structured organization to such knowledge, including approaches based on machine learning techniques; evaluation methods for ontology learning, aiming at defining procedures and metrics for a quantitative evaluation of the ontology learning task; and finally application scenarios that make ontology learning a challenging area in the context of real applications such as bio-informatics. According to the three perspectives mentioned above, the book is divided into three sections, each including a selection of papers addressing respectively the methods, the applications and the evaluation of ontology learning approaches. However, all selected papers pay considerably attention to the evaluation perspective, as this was a central topic of the ECAI 2004 workshop out of which most of the papers in this volume originate.

Bernardo Magnini

Papers

The PASCAL Recognising Textual Entailment Challenge

The Third PASCAL Recognizing Textual Entailment Challenge

Ontology Learning from Text: Methods, Evaluation and Applications

Integrating Subject Field Codes into WordNet

Ontology Learning from Text: An Overview