Search

The “MultiCoNER II Multilingual Complex Named Entity Recognition” task\footnote[1]{\url{https://multiconer.github.io}} within SemEval 2023 competition focuses on identifying complex named entities (NEs), such as the titles of creative works (e.g., songs, books, movies), people with different titles (e.g., politicians, scientists, artists, athletes), different categories of products (e.g., food, drinks, clothing), and so on, in several languages. In the context of SemEval, our team, \textit{FII\_Better}, presented an exploration of a base transformer model’s capabilities regarding the task, focused more specifically on five languages (English, Spanish, Swedish, German, and Italian). We took DistilBERT (a distilled version of BERT) and BERT (Bidirectional Encoder Representations from Transformers) as two examples of basic transformer models, using DistilBERT as a baseline and BERT as the platform to create an improved model. In this process, we managed to get fair results in the chosen languages. We have managed to get moderate results in the English track (we ranked 17th out of 34), while our results in the other tracks could be further improved in the future (overall third to last).

Background: Technology allows us to predict a histopathological diagnosis, but the high costs prevent the large-scale use of these possibilities. The current liberal indication for surgery in benign thyroid conditions led to a rising frequency of incidental thyroid carcinoma, especially low-risk papillary micro-carcinomas. Methods: We selected a cohort of 148 patients with thyroid nodules by ultrasound characteristics and investigated them by fine needle aspiration cytology (FNAC)and prospective BRAF collection for 70 patients. Also, we selected 44 patients with thyroid nodules using semi-quantitative functional imaging with an oncological, 99mTc-methoxy-isobutyl-isonitrile (99mTc-MIBI) radiotracer. Results: Following a correlation with final histopathological reports in patients who underwent thyroidectomy, we introduced the results in a machine learning program (AI) in order to obtain a pattern. For semi-quantitative functional visual pattern imaging, we found a sensitivity of 33%, a specificity of 66.67%, an accuracy of 60% and a negative predicting value (NPV) of 88.6%. For the wash-out index (WOind), we found a sensitivity of 57.14%, a specificity of 50%, an accuracy of 70% and an NPV of 90.06%.The results of BRAF in FNAC included 87.50% sensitivity, 75.00% specificity, 83.33% accuracy, 75.00% NPV and 87.50% PPV. The prevalence of malignancy in our small cohort was 11.4%. Conclusions: We intend to continue combining preoperative investigations such as molecular detection in FNAC, 99mTc-MIBI scanning and AI training with the obtained results on a larger cohort. The combination of these investigations may generate an efficient and cost-effective diagnostic tool, but confirmation of the results on a larger scale is necessary.

The Delta and Omicron variants’ system was used in this research study to replicate the complex process of the SARS-CoV-2 outbreak. The generalised fractional system was designed and rigorously analysed in order to gain a comprehensive understanding of the transmission dynamics of both variants. The proposed dynamical system has heredity and memory effects, which greatly improved our ability to perceive the disease propagation dynamics. The non-singular Atangana–Baleanu fractional operator was used to forecast the current pandemic in order to meet this challenge. The Picard recursions approach can be used to ensure that the designed fractional system has at least one solution occupying the growth condition and memory function regardless of the initial conditions. The Hyers–Ulam–Rassias stability criteria were used to carry out the stability analysis of the fractional governing system of equations, and the fixed-point theory ensured the uniqueness of the solution. Additionally, the model exhibited global asymptotically stable behaviour in some conditions. The approximate behaviour of the fatal virus was investigated using an efficient and reliable fractional numerical Adams–Bashforth approach. The outcome demonstrated that there will be a significant decline in the population of those infected with the Omicron and Delta SARS-CoV-2 variants if the vaccination rate is increased (in both the symptomatic and symptomatic stages).

In recent years, deep learning has been successfully applied to medical image analysis and provided assistance to medical professionals. Machine learning is being used to offer diagnosis suggestions, identify regions of interest in images, or augment data to remove noise. Training models for such tasks require a large amount of labeled data. It is often difficult to procure such data due to the fact that these requires experts to manually label them, in addition to the privacy and legal concerns that limiting their collection. Due to this, creating self-supervision learning methods and domain-adaptation techniques dedicated to this domain is essential. This paper reviews concepts from the field of deep learning and how they have been applied to medical image analysis. We also review the current state of self-supervised learning methods and their applications to medical images. In doing so, we will also present the resource ecosystem of researchers in this field, such as datasets, evaluation methodologies, and benchmarks.

Alzheimer’s Disease (AD) is a highly prevalent condition and most of the people suffering from it receive the diagnosis late in the process. The diagnosis is currently established following an evaluation of the protein biomarkers in cerebrospinal fluid (CSF), brain imaging, cognitive tests, and the medical history of the individuals. While diagnostic tools based on CSF collections are invasive, the tools used for acquiring brain scans are expensive. Taking these into account, an early predictive system, based on Artificial Intelligence (AI) approaches, targeting the diagnosis of this condition, as well as the identification of lead biomarkers becomes an important research direction. In this survey, we review the state-of-the-art research on machine learning (ML) techniques used for the detection of AD and Mild Cognitive Impairment (MCI). We attempt to identify the most accurate and efficient diagnostic approaches, which employ ML techniques and therefore, the ones most suitable to be used in practice. Research is still ongoing to determine the best biomarkers for the task of AD classification. At the beginning of this survey, after an introductory part, we enumerate several available resources, which can be used to build ML models targeting the diagnosis and classification of AD, as well as their main characteristics. After that, we discuss the candidate markers which were used to build AI models with the best results in terms of diagnostic accuracy, as well as their limitations.

Background and objectives: cardiovascular complications (CVC) are the leading cause of death in patients with chronic kidney disease (CKD). Standard cardiovascular disease risk prediction models used in the general population are not validated in patients with CKD. We aim to systematically review the up-to-date literature on reported outcomes of computational methods such as artificial intelligence (AI) or regression-based models to predict CVC in CKD patients. Materials and methods: the electronic databases of MEDLINE/PubMed, EMBASE, and ScienceDirect were systematically searched. The risk of bias and reporting quality for each study were assessed against transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD) and the prediction model risk of bias assessment tool (PROBAST). Results: sixteen papers were included in the present systematic review: 15 non-randomized studies and 1 ongoing clinical trial. Twelve studies were found to perform AI or regression-based predictions of CVC in CKD, either through single or composite endpoints. Four studies have come up with computational solutions for other CV-related predictions in the CKD population. Conclusions: the identified studies represent palpable trends in areas of clinical promise with an encouraging present-day performance. However, there is a clear need for more extensive application of rigorous methodologies. Following the future prospective, randomized clinical trials, and thorough external validations, computational solutions will fill the gap in cardiovascular predictive tools for chronic kidney disease.

One of the aims of this paper is to present a new programming paradigm based on the new paradigms intensively used in IT industry. Implementation of these techniques can improve the quality of code through modularization, not only in terms of entities used by a program, but also in terms of states in which they pass. Another aspect followed in this paper takes into account that in the development of software applications, the transition from the design to the source code is a very expensive step in terms of effort and time spent. Diagrams can hide very important details for simplicity of understanding, which can lead to incorrect or incomplete implementations. To improve this process communicative automaton based programming comes with an intermediate step. We will see how it goes after creating modeling diagrams to communicative automata and then to writing code for each of them. We show how the transition from one step to another is much easier and intuitive.

In this paper we present an image retrieval system created within the research project MUCKE (Multimedia and User Credibility Knowledge Extraction), a CHIST-ERA research project where UAIC{\footnote{"Alexandru Ioan Cuza" University of Iasi}} is one of the partners{\footnote{Together with Technical University from Wienna, Austria, CEA-LIST Institute from Paris, France and BILKENT University from Ankara, Turkey}}. Our discussion in this work will focus mainly on components that are part of our image retrieval system proposed in MUCKE, and we present the work done by the UAIC group. MUCKE incorporates modules for processing multimedia content in different modes and languages (like English, French, German and Romanian) and UAIC is responsible with text processing tasks (for Romanian and English). One of the problems addressed by our work is related to search results diversification. In order to solve this problem, we first process the user queries in both languages and secondly, we create clusters of similar images.

This paper approaches and tries to offer a solution to a modern day concern: relevant information retrieval and surveillance through the internet. The increasing volume of information that exists on web sites, forums or personal web pages, makes the process of searching for information that is relevant for us to become very complex and time consuming. In addition to their success, social networks like Twitter, MySpace, Facebook, Flickr have more and more users with common interests, and their gathered information has an increasing volume. In such social networks starting from a topic, users can freely express their opinions, add links or relevant photos. The system that we have built uses information collected from the Internet and offers users an easier way to find out positive and negative opinions about a topic. Information is searched on web pages (we prefer blogs, forums or users comments) and on Twitter. Identifying and classifying opinions is done by identifying some emotional triggers and by calculating some valences related to the context in which they appear.

In the last years the computational Grids have become an important research area in large-scale scientific and engineering research. Our approach is based on Peer-to-peer (P2P) networks, which are recognized as one of most used architectures in order to achieve scalability in key components of Grid systems. The main scope in using of a computational Grid was to improve the computational speed of systems that solve complex problems from Natural Language processing field. We will see how can be implemented a computational Grid using the P2P model, and how can be used SMB protocol for file transfer. After that we will see how we can use this computational Grid, in order to improve the computational speed of a system used in RTE competition [1], a new complex challenge from Natural Language processing field.

The popularity of distributed file systems continues to grow in last years. The reasons they are preferred over traditional centralized systems include fault tolerance, availability, scalability and performance. In this paper, we propose a framework for analyzing peer-to-peer content distributed technologies and their applications in the cooperative solving of combinatorial optimization problems. Our approach, which follows the Content Addressable Network model, is scalable, fault-tolerant and self-organizing; we improved also load distribution at the insertion and deletion of nodes. We use this network for the classical "graph coloring" problem, in order to reduce the computational time for its cooperative solving.