Image captioning applied to biomedical images can assist and accelerate the diagnosis process followed by clinicians. This article is first survey of image captioning, discussing datasets, evaluation measures, state art methods. Additionally, we suggest two baselines, a weak stronger one; latter outperforms all current systems on one datasets.

Abstract Diagnostic captioning (DC) concerns the automatic generation of a diagnostic text from set medical images patient collected during an examination. DC can assist inexperienced physicians, reducing clinical errors. It also help experienced physicians produce reports faster. Following advances deep learning, especially in generic image captioning, has recently attracted more attention, leading to several systems and datasets. This article is extensive overview DC. presents relevant...

In the weakly supervised learning paradigm, labeling functions automatically assign heuristic, often noisy, labels to data samples. this work, we provide a method for from weak by separating two types of complementary information associated with functions: related target label and specific one function only. Both are reflected different degrees all labeled instances. contrast previous works that aimed at correcting or removing wrongly instances, learn branched deep model uses as-is, but...

Abstract Automatic correction of errors in Handwritten Text Recognition (HTR) output poses persistent challenges yet to be fully resolved. In this study, we introduce a shared task aimed at addressing challenge, which attracted 271 submissions, yielding only handful successful approaches. This paper presents the datasets, most effective methods, and an experimental analysis error-correcting HTRed manuscripts papyri Byzantine Greek, language that followed Classical preceded Modern Greek. By...

Abstract The automated correction of errors in the Handwritten Text Recognition (HTR) output can be challenging and is far from solved. To address this challenge, we set up a shared task on AIcrowd that received 271 submissions, which very few succeed. This paper presents datasets, best methods, experimental analysis error-correcting HTRed manuscripts papyri Byzantine Greek, language followed Classical preceded Modern Greek. By using recognised transcribed data seven centuries, two...

Abstract Dating papyri accurately is crucial not only to editing their texts but also for our understanding of palaeography and the history writing, ancient scholarship, material culture, networks in antiquity, etc. Most manuscripts offer little evidence regarding time production, forcing papyrologists date them on palaeographical grounds, a method often criticized its subjectivity. In this work, with data obtained from \href{https://www.baylor.edu/classics/index.php?id=958430}{Collaborative...

Abstract Objective The study sought to assist practitioners in identifying and prioritizing radiography exams that are more likely contain abnormalities, provide them with a diagnosis order manage heavy workload efficiently (eg, during pandemic) or avoid mistakes due tiredness. Materials Methods This article introduces RTEx, novel framework for (1) ranking based on their probability be abnormal, (2) generating abnormality tags abnormal exams, (3) providing diagnostic explanation natural...

Abstract Automatic correction of errors in Handwritten Text Recognition (HTR) output poses persistent challenges yet to be fully resolved. In this study, we introduce a shared task aimed at addressing challenge, which attracted 271 submissions, yielding only handful successful approaches. This paper presents the datasets, most effective methods, and an experimental analysis error-correcting HTRed manuscripts papyri Byzantine Greek, language that followed Classical preceded Modern Greek. By...

Training data memorization in language models impacts model capability (generalization) and safety (privacy risk). This paper focuses on analyzing prompts' impact detecting the of 6 masked model-based named entity recognition models. Specifically, we employ a diverse set 400 automatically generated prompts, pairwise dataset where each pair consists one person's name from training another out set. A prompt completed with serves as input for getting model's confidence predicting this name....

This paper presents the first study for temporal relation extraction in a zero-shot setting focusing on biomedical text. We employ two types of prompts and five LLMs (GPT-3.5, Mixtral, Llama 2, Gemma, PMC-LLaMA) to obtain responses about relations between events. Our experiments demonstrate that struggle performing worse than fine-tuned specialized models terms F1 score, showing this is challenging task LLMs. further contribute novel comprehensive analysis by calculating consistency scores...

Image captioning applied to biomedical images can assist and accelerate the diagnosis process followed by clinicians. This article is first survey of image captioning, discussing datasets, evaluation measures, state art methods. Additionally, we suggest two baselines, a weak stronger one; latter outperforms all current systems on one datasets.

The Shared Task on Hateful Memes is a challenge that aims at the detection of hateful content in memes by inviting implementation systems understand memes, potentially combining image and textual information. consists three tasks: hate, protected category attack type. first binary classification task, while other two are multi-label tasks. Our participation included text-based BERT baseline (TxtBERT), same but adding information from (ImgBERT), neural retrieval approaches. We also...

Memes are a popular form of communicating trends and ideas in social media on the internet general, combining modalities images text. They can express humor sarcasm but also have offensive content. Analyzing classifying memes automatically is challenging since their interpretation relies understanding visual elements, language, background knowledge. Thus, it important to meaningfully represent these sources interaction between them order classify meme as whole. In this work, we propose use...

Handwritten text recognition (HTR) yields textual output that comprises errors, which are considerably more compared to of recognised printed (OCRed) text. Post-correcting methods can eliminate such errors but may also introduce errors. In this study, we investigate the issues arising from reality in Byzantine Greek. We properties texts lead post-correction systems adversarial behaviour and experiment with classification learn detect incorrect output. A large masked language model,...

This paper introduces RTEx, a novel methodology for a) ranking radiography exams based on their probability to contain an abnormality, b) generating abnormality tags abnormal exams, and c) providing diagnostic explanation in natural language each exam. The task of is important first step practitioners who want identify prioritize those that are more likely abnormalities, example, avoid mistakes due tiredness or manage heavy workload (e.g., during pandemic). We used two publicly available...