Structuring latent representations in a hierarchical manner enables models to learn patterns at multiple levels of abstraction. However, most prevalent image understanding models focus on visual similarity, and learning visual hierarchies is relatively unexplored. In this work, for the first time, we introduce a learning paradigm that can encode user-defined multi-level complex visual hierarchies in hyperbolic space without requiring explicit hierarchical labels. As a concrete example, first, we define a part-based image hierarchy using object-level annotations within and across images. Then, we introduce an approach to enforce the hierarchy using contrastive loss with pairwise entailment metrics. Finally, we discuss new evaluation metrics to effectively measure hierarchical image retrieval. Encoding these complex relationships ensures that the learned representations capture semantic and structural information that transcends mere visual similarity. Experiments in part-based image retrieval show significant improvements in hierarchical retrieval tasks, demonstrating the capability of our model in capturing visual hierarchies.

The prevalence of neurological disorders poses a challenge to modern healthcare, requiring advancements in diagnostic and prognostic methodologies. This study introduces a deep generative model that retroactively reconstructs magnetic resonance imaging data of human brains into their longitudinal counterparts, creating valuable methods for facilitating meticulous analyses of disease progression. The lack of imaging data on healthy individuals compared to those with brain degenerative disorders, coupled with the time-sensitive nature of some diseases, makes their early diagnosis and effective treatment complex. We demonstrate the model's efficacy in generating anatomically accurate brain scans to aid in comprehending the dynamic nature of brain pathology, as evidenced by our mixed-method study: Our quantitative evaluation resulted in an outstanding Frechet Inception Distance score of 5.801 and competitive performance in other key metrics compared to other state-of-the-art inpainting models. Our qualitative evaluation, conducted by two general radiologists and two neuroradiologists, yielded a Discrimination Success Rate of 51.67%, indicating the model's success in generating realistic images. By integrating this methodology into clinical practice, we anticipate enhanced patient outcomes by personalizing precision medicine and emphasizing preventive strategies such as early and tailored therapeutic interventions.

Estimating the performance of a machine learning system is a longstanding challenge in artificial intelligence research. Today, this challenge is especially relevant given the emergence of systems which appear to increasingly outperform human beings. In some cases, this "superhuman" performance is readily demonstrated; for example by defeating legendary human players in traditional two player games. On the other hand, it can be challenging to evaluate classification models that potentially surpass human performance. Indeed, human annotations are often treated as a ground truth, which implicitly assumes the superiority of the human over any models trained on human annotations. In reality, human annotators can make mistakes and be subjective. Evaluating the performance with respect to a genuine oracle may be more objective and reliable, even when querying the oracle is expensive or impossible. In this paper, we first raise the challenge of evaluating the performance of both humans and models with respect to an oracle which is unobserved. We develop a theory for estimating the accuracy compared to the oracle, using only imperfect human annotations for reference. Our analysis provides a simple recipe for detecting and certifying superhuman performance in this setting, which we believe will assist in understanding the stage of current research on classification. We validate the convergence of the bounds and the assumptions of our theory on carefully designed toy experiments with known oracles. Moreover, we demonstrate the utility of our theory by meta-analyzing large-scale natural language processing tasks, for which an oracle does not exist, and show that under our assumptions a number of models from recent years are with high probability superhuman.

This paper addresses a key question in current machine learning research: if we believe that a model's predictions might be better than those given by human experts, how can we (humans) verify these beliefs? In some cases, this ``superhuman'' performance is readily demonstrated; for example by defeating top-tier human players in traditional two player games. On the other hand, it can be challenging to evaluate classification models that potentially surpass human performance. Indeed, human annotations are often treated as a ground truth, which implicitly assumes the superiority of the human over any models trained on human annotations. In reality, human annotators are subjective and can make mistakes. Evaluating the performance with respect to a genuine oracle is more objective and reliable, even when querying the oracle is more expensive or sometimes impossible. In this paper, we first raise the challenge of evaluating the performance of both humans and models with respect to an oracle which is $\textit{unobserved}$. We develop a theory for estimating the accuracy compared to the oracle, using only imperfect human annotations for reference. Our analysis provides an executable recipe for detecting and certifying superhuman performance in this setting, which we believe will assist in understanding the stage of current research on classification. We validate the convergence of the bounds and the assumptions of our theory on carefully designed toy experiments with known oracles. Moreover, we demonstrate the utility of our theory by meta-analyzing large-scale natural language processing tasks, for which an oracle does not exist, and show that under our mild assumptions a number of models from recent years have already achieved superhuman performance with high probability---suggesting that our new oracle based performance evaluation metrics are overdue as an alternative to the widely used accuracy metrics that are naively based on imperfect human annotations.

Semi-supervised semantic segmentation methods use a small amount of clean pixel-level annotations to guide the interpretation of a larger quantity of unlabelled image data. The challenges of providing pixel-accurate annotations at scale mean that the labels are typically noisy, and this contaminates the final results. In this work, we propose an approach that is robust to label noise in the annotated data. The method uses two diverse learning groups with different network architectures to effectively handle both label noise and unlabelled images. Each learning group consists of a teacher network, a student network and a novel filter module. The filter module of each learning group utilizes pixel-level features from the teacher network to detect incorrectly labelled pixels. To reduce confirmation bias, we employ the labels cleaned by the filter module from one learning group to train the other learning group. Experimental results on two different benchmarks and settings demonstrate the superiority of our method over state-of-the-art approaches.

This work studies the task of glossification, of which the aim is to em transcribe natural spoken language sentences for the Deaf (hard-of-hearing) community to ordered sign language glosses. Previous sequence-to-sequence language models trained with paired sentence-gloss data often fail to capture the rich connections between the two distinct languages, leading to unsatisfactory transcriptions. We observe that despite different grammars, glosses effectively simplify sentences for the ease of deaf communication, while sharing a large portion of vocabulary with sentences. This has motivated us to implement glossification by executing a collection of editing actions, e.g. word addition, deletion, and copying, called editing programs, on their natural spoken language counterparts. Specifically, we design a new neural agent that learns to synthesize and execute editing programs, conditioned on sentence contexts and partial editing results. The agent is trained to imitate minimal editing programs, while exploring more widely the program space via policy gradients to optimize sequence-wise transcription quality. Results show that our approach outperforms previous glossification models by a large margin, improving the BLEU-4 score from 16.45 to 18.89 on RWTH-PHOENIX-WEATHER-2014T and from 18.38 to 21.30 on CSL-Daily.

A multi-language dictionary is a fundamental tool for language learning, allowing the learner to look up unfamiliar words. Searching an unrecognized word in the dictionary does not usually require deep knowledge of the target language. However, this is not true for sign language, where gestural elements preclude this type of easy lookup. This paper introduces GlossFinder, an online tool supporting 2, 000 signs to assist language learners in determining the meaning of given signs. Unlike alternative systems of complex inputs, our system requires only that learners imitate the sign in front of a standard webcam. A user study conducted among sign language speakers of varying ability compared our system against existing alternatives and the interviews indicated a clear preference for our new system. This implies that GlossFinder can lower the barrier in sign language learning by addressing the common problem of sign finding and make it accessible to the wider community.

In this paper, we provide an overview of the ninth annual edition of the CLEF eHealth evaluation lab. CLEF eHealth 2021 continues our evaluation resource building efforts around the easing and support of patients, their next-of-kins, health care professionals, and health scientists in understanding, accessing, and authoring electronic health information in a multilingual setting. The 2021 lab offered two tasks: Task 1 on multilingual Information Extraction (IE), this year extending to a corpus of Spanish radiology reports; and Task 2 on Consumer Health Search (CHS) that builds on the previous year's Information Retrieval (IR) tasks. In total, 11 teams took part in these tasks (7 in Task 1 on IE and 4 in Task 2 on IR). Herein, we describe the resources created for these tasks and the evaluation methodology adopted, and we provide a brief summary of the participants of this year's challenges as well as the results obtained. As in previous years, the organizers have made data, tools, and more specific overview papers associated with the lab tasks available for future research and development.

The Bidirectional Encoder Representations from Transformers (BERT) model has achieved the state-of-the-art performance for many natural language processing (NLP) tasks. Yet, limited research has been contributed to studying its effectiveness when the target domain is shifted from the pre-training corpora, for example, for biomedical or clinical NLP applications. In this paper, we applied it to a widely studied a hospital information extraction (IE) task and analyzed its performance under the transfer learning setting. Our application became the new state-of-the-art result by a clear margin, compared with a range of existing IE models. Specifically, on this nursing handover data set, the macro-average F1 score from our model was 0.438, whilst the previous best deep learning models had 0.416. In conclusion, we showed that BERT based pre-training models can be transferred to health-related documents under mild conditions and with a proper fine-tuning process.

Leakage of personal information in conversations raises serious privacy concerns. Malicious people or bots could pry into sensitive personal information of vulnerable people, such as juveniles, through conversations with them or their digital personal assistants. To address the problem, we present a privacy-leakage warning system that monitors conversations in social media and intercepts the outgoing text messages from a user or a digital assistant, if they impose potential privacy leakage risks. Such messages are redirected to authorized users for approval, before they are sent out. We demonstrate how our system is deployed and used on a social media conversation platform, e.g., Facebook Messenger.

Motivated by the ever increasing difficulties faced by laypeople in retrieving and digesting valid and relevant information to make health-centred decisions, the CLEF eHealth lab series has offered shared tasks to the community in the fields of Information Extraction (IE), management, and Information Retrieval (IR) since 2013. These tasks have attracted large participation and led to statistically significant improvements in processing quality. In 2021, CLEF eHealth is calling for participants to contribute to the following two tasks: Task 1 on IE focuses on IE from noisy text. Participants will identify and classify Named Entities in written ultrasonography reports, containing misspellings and inconsistencies, from a major public hospital in Argentina. Identified entities will then have to be classified, which can be very challenging as it requires to handle lexical variations. Task 2 is a novel extension of the most popular and established task on consumer health search (CHS), aiming at retrieving relevant, understandable, and credible information for patients and their next-of-kins. In this paper we describe recent advances in the fields of IE and IR, and the subsequent offerings of this years CLEF eHealth lab challenges.

Well-annotated datasets, as shown in recent top studies, are becoming more important for researchers than ever before in supervised machine learning (ML). However, the dataset annotation process and its related human labor costs remain overlooked. In this work, we analyze the relationship between the annotation granularity and ML performance in sequence labeling, using clinical records from nursing shift-change handover. We first study a model derived from textual language features alone, without additional information based on nursing knowledge. We find that this sequence tagger performs well in most categories under this granularity. Then, we further include the additional manual annotations by a nurse, and find the sequence tagging performance remaining nearly the same. Finally, we give a guideline and reference to the community arguing it is not necessary and even not recommended to annotate in detailed granularity because of a low Return on Investment. Therefore we recommend emphasizing other features, like textual knowledge, for researchers and practitioners as a cost-effective source for increasing the sequence labeling performance.

Video deblurring models exploit consecutive frames to remove blurs from camera shakes and object motions. In order to utilize neighboring sharp patches, typical methods rely mainly on homography or optical flows to spatially align neighboring blurry frames. However, such explicit approaches are less effective in the presence of fast motions with large pixel displacements. In this work, we propose a novel implicit method to learn spatial correspondence among blurry frames in the feature space. To construct distant pixel correspondences, our model builds a correlation volume pyramid among all the pixel-pairs between neighboring frames. To enhance the features of the reference frame, we design a correlative aggregation module that maximizes the pixel-pair correlations with its neighbors based on the volume pyramid. Finally, we feed the aggregated features into a reconstruction module to obtain the restored frame. We design a generative adversarial paradigm to optimize the model progressively. Our proposed method is evaluated on the widely-adopted DVD dataset, along with a newly collected High-Frame-Rate (1000 fps) Dataset for Video Deblurring (HFR-DVD). Quantitative and qualitative experiments show that our model performs favorably on both datasets against previous state-of-the-art methods, confirming the benefit of modeling all-range spatial correspondence for video deblurring.

This paper details materials, methods, results, and analyses of the Consumer Health Search Task of the CLEF eHealth 2021 Evaluation Lab. This task investigates the effectiveness of information retrieval (IR) approaches in providing access to medical information to laypeople. For this a TREC -style evaluation methodology was applied: a shared collection of documents and queries is distributed, participants’ runs received, relevance assessments generated, and participants’ submissions evaluated. The task generated a new representative web corpus including web pages acquired from a 2021 CommonCrawl and social media content from Twitter and Reddit, along with a new collection of 55 manually generated layperson medical queries and their respective credibility, understandability, and topicality assessments for returned documents. This year’s task focused on three subtask: (i) ad-hoc IR, (ii) weakly supervised IR, and (iii) document credibility prediction. In total, 15 runs were submitted to the three subtasks: eight addressed the ad-hoc IR task, three the weakly supervised IR challenge, and 4 the document credibility prediction challenge. As in previous years, the organizers have made data and tools associated with the task available for future research and development.

In the time of COVID-19, people has rigid demand on seeking accurate health information from internet. However, health information searching is still a challenge in information retrieval (IR) technology and real-world medical informatics applications. To improve healthcare, many medical search engines have been built based on natural language processing, machine learning, and other computational techniques to address related IR problems, but the result have not always been satisfactory, especially to laypersons, because those information access tools cannot, for example, retrieve documents with high relevance and support people’s need.On the country, models based on IR produce a relatively ideal performance on relevance, readability, and credibility.Therefore, IR technology has become popular and widespread. The work addressed three aims: First, I implemented a BM25-based IR model with a query expansion using the REINFORCE algorithm on a benchmark dataset, called clefehealth2018_B released as part of the CLEF eHealth Lab by the Conference and Labs of the Evaluation Forum. Second, I compared my method with the baseline. Third, as one of the organisers of this year’s CLEF eHealth IR Task, I set up a website-based evaluation tool for manually labelling submitted documents from participants. In the experiments, the new method achieved a better performance than the baseline with the best Precision@10 and NDCG@10 scores being 0.534 and 0.492, respectively. The results gave evidence of the proposed method being able to return more relevant documents and the primary information extraction operation in the data pre-processing improved the model performance further. The novelty of the research was training the REINFORCE algorithm based reformulator for query expansion. Furthermore, the detailed comparisons between different searching engines and various settings were illustrated. Finally, this work resulted in a new IR idea to further study on the benchmark dataset using a reinforcement learning method to expand queries.

Sign language translation (SLT) aims to interpret sign video sequences into text-based natural language sentences. Sign videos consist of continuous sequences of sign gestures with no clear boundaries in between. Existing SLT models usually represent sign visual features in a frame-wise manner so as to avoid needing to explicitly segmenting the videos into isolated signs. However, these methods neglect the temporal information of signs and lead to substantial ambiguity in translation. In this paper, we explore the temporal semantic structures of sign videos to learn more discriminative features. To this end, we first present a novel sign video segment representation which takes into account multiple temporal granularities, thus alleviating the need for accurate video segmentation. Taking advantage of the proposed segment representation, we develop a novel hierarchical sign video feature learning method via a temporal semantic pyramid network, called TSPNet. Specifically, TSPNet introduces an inter-scale attention to evaluate and enhance local semantic consistency of sign segments and an intra-scale attention to resolve semantic ambiguity by using non-local video context. Experiments show that our TSPNet outperforms the state-of-the-art with significant improvements on the BLEU score (from 9.58 to 13.41) and ROUGE score (from 31.80 to 34.96) on the largest commonly-used SLT dataset. Our implementation is available at https://github.com/verashira/TSPNet.

Word-level sign language recognition (WSLR) is a fundamental task in sign language interpretation. It requires models to recognize isolated sign words from videos. However, annotating WSLR data needs expert knowledge, thus limiting WSLR dataset acquisition. On the contrary, there are abundant subtitled sign news videos on the internet. Since these videos have no word-level annotation and exhibit a large domain gap from isolated signs, they cannot be directly used for training WSLR models. We observe that despite the existence of a large domain gap, isolated and news signs share the same visual concepts, such as hand gestures and body movements. Motivated by this observation, we propose a novel method that learns domain-invariant visual concepts and fertilizes WSLR models by transferring knowledge of subtitled news sign to them. To this end, we extract news signs using a base WSLR model, and then design a classifier jointly trained on news and isolated signs to coarsely align these two domain features. In order to learn domain-invariant features within each class and suppress domain-specific features, our method further resorts to an external memory to store the class centroids of the aligned news signs. We then design a temporal attention based on the learnt descriptor to improve recognition performance. Experimental results on standard WSLR datasets show that our method outperforms previous state-of-the-art methods significantly. We also demonstrate the effectiveness of our method on automatically localizing signs from sign news, achieving 28.1 for AP@0.5.

In this paper, we provide an overview of the CLEF eHealth Task 2 on Information Retrieval (IR), organized as part of the eighth annual edition of the CLEF eHealth evaluation lab by the Conference and Labs of the Evaluation Forum. Its aim was to address laypeople’s difficulties in retrieving and digesting valid and relevant information, in their preferred language, to make health-centred decisions. The task was a novel extension of the most popular and established task in CLEF eHealth on Consumer Health Search (CHS), which makes responses to spoken ad-hoc queries. In total, five submissions were made to its two subtasks; three addressed the ad-hoc IR task on text data and two considered the spoken queries. Herein, we describe the resources created for the task and evaluation methodology adopted. We also summarize lab submissions and results. As in previous years, organizers have made data, methods, and tools associated with the lab tasks available for future research and development. Copyright © 2020 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). CLEF 2020, 22-25 September 2020, Thessaloniki, Greece. ? With equal contribution, LG & HS were the co-leaders of the task. LK, ZL, GP, GGS, MV, and CX were the task co-organizers and contributors to the evaluation conceptualization, dataset creation, assessments, and measurements.

In this paper, we provide an overview of the eight annual edition of the Conference and Labs of the Evaluation Forum (CLEF) eHealth evaluation lab. The Conference and Labs of the Evaluation Forum (CLEF) eHealth 2020 continues our development of evaluation tasks and resources since 2012 to address laypeople’s difficulties to retrieve and digest valid and relevant information in their preferred language to make health-centred decisions. This year’s lab advertised two tasks. Task 1 on Information Extraction (IE) was new and focused on automatic clinical coding of diagnosis and procedure the tenth revision of the International Statistical Classification of Diseases and Related Health Problems (ICD10) codes as well as finding the corresponding evidence text snippets for clinical case documents in Spanish. Task 2 on Information Retrieval (IR) was a novel extension of the most popular and established task in the Conference and Labs of the Evaluation Forum (CLEF) eHealth on Consumer Health Search (CHS). In total 55 submissions were made to these tasks. Herein, we describe the resources created for the two tasks and evaluation methodology adopted. We also summarize lab submissions and results. As in previous years, the organizers have made data and tools associated with the lab tasks available for future research and development. The ongoing substantial community interest in the tasks and their resources has led to the Conference and Labs of the Evaluation Forum (CLEF) eHealth maturing as a primary venue for all interdisciplinary actors of the ecosystem for producing, processing, and consuming electronic health information.

In this paper, we describe ALTER, an auxiliary text rewriting tool that facilitates the rewriting process for natural language generation tasks, such as paraphrasing, text simplification, fairness-aware text rewriting, and text style transfer. Our tool is characterized by two features, i) recording of word-level revision histories and ii) flexible auxiliary edit support and feedback to annotators. The text rewriting assist and traceable rewriting history are potentially beneficial to the future research of natural language generation.