PARE: A Simple and Strong Baseline for Monolingual and Multilingual Distantly Supervised Relation Extraction

Neural models for distantly supervised relation extraction (DS-RE) encode each sentence in an entity-pair bag separately. These are then aggregated for bag-level relation prediction. Since, at encoding time, these approaches do not allow information to flow from other sentences in the bag, we believe that they do not utilize the available bag data to the fullest. In response, we explore a simple baseline approach (PARE) in which all sentences of a bag are concatenated into a passage of sentences, and encoded jointly using BERT. The contextual embeddings of tokens are aggregated using attention with the candidate relation as query -- this summary of whole passage predicts the candidate relation. We find that our simple baseline solution outperforms existing state-of-the-art DS-RE models in both monolingual and multilingual DS-RE datasets.

The role of social networks during natural disasters is becoming crucial to share relevant information and coordinate relief actions. With the reach of the social networks, any user around the world has the possibility of interact in crisis-events as these unfold. A large part of the information posted during a disaster uses the native language where the disaster occurred. However, there are also users from other parts of the world who can comment about the event, often in another language. In this work, we conducted a study of crisis-related tweets about the earthquake that occurred in Ecuador in April 2016. To that end, we introduce a new annotated dataset in both Spanish and English languages with approximately 8K tweets; half of them belong to conversations. We evaluate several neural architectures to identify crisis-related tweets in a multi-lingual setting, and we found that deep contextual multi-lingual embeddings outperform other strong baseline models. We then explore the type of conversations that occur from the perspective of different languages. The results show that certain types of conversations occur more in the native language and others in a foreign language. Conversations from foreign countries seek to gather situation awareness and give emotional support, while in the affected country the conversations aim mainly to humanitarian aid.

Abstract3D multi-object tracking (MOT) has witnessed numerous novel benchmarks and approaches in recent years, especially those under the “tracking-by-detection” paradigm. Despite their progress and usefulness, an in-depth analysis of their strengths and weaknesses is not yet available. In this paper, we summarize current 3D MOT methods into a unified framework by decomposing them into four constituent parts: pre-processing of detection, association, motion model, and life cycle management. We then ascribe the failure cases of existing algorithms to each component and investigate them in detail. Based on the analyses, we propose corresponding improvements which lead to a strong yet simple baseline: SimpleTrack. Comprehensive experimental results on Waymo Open Dataset and nuScenes demonstrate that our final method could achieve new state-of-the-art results with minor modifications. Furthermore, we take additional steps and rethink whether current benchmarks authentically reflect the ability of algorithms for real-world challenges. We delve into the details of existing benchmarks and find some intriguing facts. Finally, we analyze the distribution and causes of remaining failures in SimpleTrack and propose future directions for 3D MOT. Our code is at https://github.com/tusen-ai/SimpleTrack.Keywords3D multi-object trackingAutonomous driving

BasicTAD: An astounding RGB-Only baseline for temporal action detection

This study explores a simple but strong baseline for person re-identification (ReID). Person ReID with deep neural networks has progressed and achieved high performance in recent years. However, many state-of-the-art methods design complex network structures and concatenate multi-branch features. In the literature, some effective training tricks briefly appear in several papers or source codes. The present study collects and evaluates these effective training tricks in person ReID. By combining these tricks, the model achieves 94.5% rank-1 and 85.9% mean average precision on Market1501 with only using the global features of ResNet50. The performance surpasses all existing global- and part-based baselines in person ReID. We propose a novel neck structure named as batch normalization neck (BNNeck). BNNeck adds a batch normalization layer after global pooling layer to separate metric and classification losses into two different feature spaces because we observe they are inconsistent in one embedding space. Extended experiments show that BNNeck can boost the baseline, and our baseline can improve the performance of existing state-of-the-art methods. Our codes and models are available at: https://github.com/michuanhaohao/reid-strong-baseline.

The World Wide Web includes semantic relations of numerous types that exist among different entities. Extracting the relations that exist between two entities is an important step in various Web-related tasks such as information retrieval (IR), information extraction, and social network extraction. A supervised relation extraction system that is trained to extract a particular relation type (source relation) might not accurately extract a new type of a relation (target relation) for which it has not been trained. However, it is costly to create training data manually for every new relation type that one might want to extract. We propose a method to adapt an existing relation extraction system to extract new relation types with minimum supervision. Our proposed method comprises two stages: learning a lower dimensional projection between different relations, and learning a relational classifier for the target relation type with instance sampling. First, to represent a semantic relation that exists between two entities, we extract lexical and syntactic patterns from contexts in which those two entities co-occur. Then, we construct a bipartite graph between relation-specific (RS) and relation-independent (RI) patterns. Spectral clustering is performed on the bipartite graph to compute a lower dimensional projection. Second, we train a classifier for the target relation type using a small number of labeled instances. To account for the lack of target relation training instances, we present a one-sided under sampling method. We evaluate the proposed method using a data set that contains 2,000 instances for 20 different relation types. Our experimental results show that the proposed method achieves a statistically significant macroaverage F-score of 62.77. Moreover, the proposed method outperforms numerous baselines and a previously proposed weakly supervised relation extraction method.

Distantly supervised relation extraction is an automatically annotating method for large corpora by classifying a bound of sentences with two same entities and the relation. Recent works exploit sound performance by adopting contrastive learning to efficiently obtain instance representations under the multi-instance learning framework. Though these methods weaken the impact of noisy labels, it ignores the long-tail distribution problem in distantly supervised sets and fails to capture the mutual information of different parts. We are thus motivated to tackle these issues and establishing a long-tail awareness contrastive learning method for efficiently utilizing the long-tail data. Our model treats major and tail parts differently by adopting hyper-augmentation strategies. Moreover, the model provides various views by constructing novel positive and negative pairs in contrastive learning for gaining a better representation between different parts. The experimental results on the NYT10 dataset demonstrate our model surpasses the existing SOTA by more than 2.61% AUC score on relation extraction. In manual evaluation datasets including NYT10m and Wiki20m, our method obtains competitive results by achieving 59.42% and 79.19% AUC scores on relation extraction, respectively. Extensive discussions further confirm the effectiveness of our approach.

Image-based visual-language (I-VL) pre-training has shown great success for learning joint visual-textual representations from large-scale web data, revealing remarkable ability for “zero-shot” generalisation. This paper presents a simple but strong baseline to efficiently adapt the pre-trained I-VL model for video understanding tasks, with minimal training. Specifically, we propose to optimise a few random vectors, termed as “continuous prompt vectors”, that convert video-related tasks into the same format as the pre-training objectives. In addition, to bridge the gap between static images and videos, temporal information is encoded with lightweight Transformers stacking on top of frame-wise visual features. Experimentally, we conduct extensive ablation studies to analyse the critical components. On ten public benchmarks of action recognition, action localisation, and text-video retrieval, across closed-set, few-shot, and zero-shot scenarios, we achieve competitive or state-of-the-art performance to existing methods, despite optimising significantly fewer parameters. Due to space limitation, we refer the readers to the arXiv version at https://arxiv.org/abs/2112.04478.

Semi-supervised learning (SSL) methods effectively leverage unlabeled data to improve model generalization. However, SSL models often underperform in open-set scenarios, where unlabeled data contain outliers from novel categories that do not appear in the labeled set. In this paper, we study the challenging and realistic open-set SSL setting, where the goal is to both correctly classify inliers and to detect outliers. Intuitively, the inlier classifier should be trained on inlier data only. However, we find that inlier classification performance can be largely improved by incorporating high-confidence pseudo-labeled data, regardless of whether they are inliers or outliers. Also, we propose to utilize non-linear transformations to separate the features used for inlier classification and outlier detection in the multi-task learning framework, preventing adverse effects between them. Additionally, we introduce pseudo-negative mining, which further boosts outlier detection performance. The three ingredients lead to what we call Simple but Strong Baseline (SSB) for open-set SSL. In experiments, SSB greatly improves both inlier classification and outlier detection performance, outperforming existing methods by a large margin. Our code will be released at https://github.com/YUE-FAN/SSB.

Video Instance Segmentation (VIS) is a task that simultaneously requires classification, segmentation, and instance association in a video. Recent VIS approaches rely on sophisticated pipelines to achieve this goal, including RoI-related operations or 3D convolutions. In contrast, we present a simple and efficient single-stage VIS framework based on the instance segmentation method CondInst by adding an extra tracking head. To improve instance association accuracy, a novel bi-directional spatio-temporal contrastive learning strategy for tracking embedding across frames is proposed. Moreover, an instance-wise temporal consistency scheme is utilized to produce temporally coherent results. Experiments conducted on the YouTube-VIS-2019, YouTube-VIS-2021, and OVIS-2021 datasets validate the effectiveness and efficiency of the proposed method. We hope the proposed framework can serve as a simple and strong baseline for other instance-level video association tasks.

Relation extraction is an important task for understanding relationships between entities, building knowledge graphs, and facilitating knowledge discovery. Pre-trained models can be fine-tuned for relation extraction if a substantial amount of labeled data is available. However, acquiring extensive labeled data is generally challenging. Semi-supervised techniques for low-resource relation extraction, such as self-training, offer a promising solution by leveraging both limited labeled data and vast unlabeled data to mitigate this challenge. Traditional self-training methods use a teacher-student framework, where a student is iteratively trained with pseudo-labels generated by the teacher. This may lead to noisy pseudo-labels and impact performance. To address this limitation, we introduce a new model called RE-AUM-LLM that generates high-quality pseudo-labels using self-training combined with Area Under the Margin (AUM) and Large Language Models (LLMs), such as Llama 3.1. Experimental results on two benchmark datasets show that the proposed approach achieves state-of-the-art results for low-resource relation extraction by comparison with several strong baselines. We will make the code publicly available to enable reproducibility and further research in this area.

The task of entity relation extraction discovers new relation facts and enables broader applications of knowledge graph. Distant supervision is widely adopted for relation extraction, which requires large amounts of texts containing entity pairs as training data. However, in some specific domains such as medical-related applications, entity pairs that have certain relations might not appear together, thus it is difficult to meet the requirement for distantly supervised relation extraction. In the light of this challenge, we propose a novel path-based model to discover new entity relation facts. Instead of finding texts for relation extraction, the proposed method extracts path-only information for entity pairs from the current knowledge graph. For each pair of entities, multiple paths can be extracted, and some of them are more useful for relation extraction than others. In order to capture this observation, we employ attention mechanism to assign different weights for different paths, which highlights the useful paths for entity relation extraction. To demonstrate the effectiveness of the proposed method, we conduct various experiments on a large-scale medical knowledge graph. Compared with the state-of-the-art relation extraction methods using the structure of knowledge graph, the proposed method significantly improves the accuracy of extracted relation facts and achieves the best performance.

Distantly Supervised Relation Extraction (DSRE) aims to automatically identify semantic relationships within large text corpora by aligning with external knowledge bases. Despite the success of current methods in automating data annotation, they introduce two main challenges: label noise and data long-tail distribution. Label noise results in inaccurate annotations, which can undermine the quality of relation extraction. The long-tail problem, on the other hand, leads to an imbalanced model that struggles to extract less frequent, long-tail relations. In this paper, we introduce a novel relation extraction framework based on multi-level hierarchical attention. This approach utilizes Graph Attention Networks (GATs) to model the hierarchical structure of the relations, capturing the semantic dependencies between relation types and generating relation embeddings that reflect the overall hierarchical framework. To improve the classification process, we incorporate a multi-level classification structure guided by hierarchical attention, which enhances the accuracy of both head and tail relation extraction. A local probability constraint is introduced to ensure coherence across the classification levels, fostering knowledge transfer from frequent to less frequent relations. Experimental evaluations on the New York Times (NYT) dataset demonstrate that our method outperforms existing baselines, particularly in the context of long-tail relation extraction, offering a comprehensive solution to the challenges of DSRE.

The goal of relation extraction is to obtain relational facts from plain text, which can benefit a variety of natural language processing tasks. To address the challenge of automatically labeling large-scale training data, a distant supervision strategy is introduced to relation extraction by heuristically aligning entity pairs in plain text with the knowledge base. Unfortunately, the method is vulnerable to the noisy label problem due to the incompletion of the exploited knowledge base. Existing works focus on the specific algorithms, but few works summarize the commonalities between different methods and the influencing factors of these denoising mechanisms. In this paper, we propose three main factors that impact the label denoising of distantly supervised relation extraction, including labeling assumption, prior knowledge and confidence level. In order to analyze how these factors influence the denoising effectiveness, we build a unified neural framework with word, sentence and label denoising modules for relation extraction. Then we conduct experiments to evaluate and compare these factors according to ten neural schemes. In addition, we discuss the typical cases of these factors and find that influential word-level prior knowledge and partial confidence for distantly supervised labels can significantly affect the denoising performance. These implicational findings can provide researchers with more insight of distantly supervised relation extraction.

Distant supervised relation extraction has been widely used to find novel relational facts from text. However, distant supervision inevitably accompanies with the wrong labelling problem, and these noisy data will substantially hurt the performance of relation extraction. To alleviate this issue, we propose a sentence-level attention-based model for relation extraction. In this model, we employ convolutional neural networks to embed the semantics of sentences. Afterwards, we build sentence-level attention over multiple instances, which is expected to dynamically reduce the weights of those noisy instances. Experimental results on real-world datasets show that, our model can make full use of all informative sentences and effectively reduce the influence of wrong labelled instances. Our model achieves significant and consistent improvements on relation extraction as compared with baselines.