Named Entity Recognition Research Articles

NSSC: a neuro-symbolic AI system for enhancing accuracy of named entity recognition and linking from oncologic clinical notes.

Accurate recognition and linking of oncologic entities in clinical notes is essential for extracting insights across cancer research, patient care, clinical decision-making, and treatment optimization. We present the Neuro-Symbolic System for Cancer (NSSC), a hybrid AI framework that integrates neurosymbolic methods with named entity recognition (NER) and entity linking (EL) to transform unstructured clinical notes into structured terms using medical vocabularies, with the Unified Medical Language System (UMLS) as a case study. NSSC was evaluated on a dataset of clinical notes from breast cancer patients, demonstrating significant improvements in the accuracy of both entity recognition and linking compared to state-of-the-art models. Specifically, NSSC achieved a 33% improvement over BioFalcon and a 58% improvement over scispaCy. By combining large language models (LLMs) with symbolic reasoning, NSSC improves the recognition and interoperability of oncologic entities, enabling seamless integration with existing biomedical knowledge. This approach marks a significant advancement in extracting meaningful information from clinical narratives, offering promising applications in cancer research and personalized patient care.

Monitoring the trade in bat taxidermy and specimens on e-commerce platforms

Wildlife trade is a major economic activity but can pose a threat to many species. Previous research has focused on the trade of bats (Chiroptera) for food and medicine but little is known about the online, international trade in bat taxidermy and specimens, which is thought to be an emerging threat. We aimed to characterise this trade by sampling two major e-commerce platforms over an 18-month period. We used an automated pipeline, whereby data were accessed programmatically; filtered using a neural-network classifier; and the locations and species mentioned in relevant listings were identified using Named Entity Recognition. We retrieved 40,412 relevant results, representing 2363 and 2116 unique listings from Etsy and eBay, respectively. Although the accuracy of listings could not be verified, most (57 %) provided species-level information. Hence, 47 bat species were listed for sale, of which 32 are not currently identified by the IUCN Red List as traded for any purpose. The highest number of listings were sold from USA, Hong Kong and the UK, whilst traded species' ranges were predominantly located in South East Asia. We provide one of the first longitudinal datasets on the online bat trade, indicating that several species are being traded at potentially high volumes, with demand from North America and Europe driving exploitation in South East Asia. This is concerning for highly-traded species which occur at low densities, such as the painted wooly bat Kerivoula picta. Our automated pipeline can be used for subsequent monitoring of this trade, or could be adapted for monitoring of trade in other taxa.

VERA-ARAB: unveiling the Arabic tweets credibility by constructing balanced news dataset for veracity analysis

The proliferation of fake news on social media platforms necessitates the development of reliable datasets for effective fake news detection and veracity analysis. In this article, we introduce a veracity dataset of Arabic tweets called “VERA-ARAB”, a pioneering large-scale dataset designed to enhance fake news detection in Arabic tweets. VERA-ARAB is a balanced, multi-domain, and multi-dialectal dataset, containing both fake and true news, meticulously verified by fact-checking experts from Misbar. Comprising approximately 20,000 tweets from 13,000 distinct users and covering 884 different claims, the dataset includes detailed information such as news text, user details, and spatiotemporal data, spanning diverse domains like sports and politics. We leveraged the X API to retrieve and structure the dataset, providing a comprehensive data dictionary to describe the raw data and conducting a thorough statistical descriptive analysis. This analysis reveals insightful patterns and distributions, visualized according to data type and nature. We also evaluated the dataset using multiple machine learning classification models, exploring various social and textual features. Our findings indicate promising results, particularly with textual features, underscoring the dataset’s potential for enhancing fake news detection. Furthermore, we outline future work aimed at expanding VERA-ARAB to establish it as a benchmark for Arabic tweets in fake news detection. We also discuss other potential applications that could leverage the VERA-ARAB dataset, emphasizing its value and versatility for advancing the field of fake news detection in Arabic social media. Potential applications include user veracity assessment, topic modeling, and named entity recognition, demonstrating the dataset's wide-ranging utility for broader research in information quality management on social media.

BiodiViz: Leveraging NER and RE for Automated Knowledge Graph Generation in Biodiversity Research

In biodiversity research, the integration of machine learning and data visualization is increasingly important for uncovering valuable insights from academic literature. This study introduces an innovative knowledge graph application, BiodiViz, designed to translate intricate text into intuitive visual representations, fostering a deeper comprehension of biodiversity relationships. BiodiViz uses the top-performing Named Entity Recognition (NER) and Relation Extraction (RE) models to automatically generate a comprehensive knowledge graph for biodiversity research. The NER model extracts and categorizes entities like organisms, phenomena, and habitats, while the RE model identifies relationships such as "have," "occur in," and "influence" from the BiodivNERE dataset (Abdelmageed et al. 2022). These entities and relationships are organized into nodes and edges within a graph. Researchers input text into BiodiViz, producing a visual knowledge graph that simplifies the analysis of complex biodiversity data, reducing manual effort and enhancing efficiency. Named Entity Recognition & Relation Extraction BiodiViz leverages advanced Bidirectional Encoder Representations from Transformers (BERT)-based Large Language Models (LLMs) (Rogers et al. 2020), fine-tuned specifically for NER and RE tasks using the BiodivNERE dataset. The fine-tuning process involved various models, including BERT (Devlin et al. 2019), ELECTRA (Clark et al. 2020), and BiodivBERT (Abdelmageed et al. 2023). These models were evaluated for performance using the results of their F1-score as the main metric, which is the harmonic mean of precision (the proportion of true positive results among all positive predictions) and recall (the proportion of true positive results among all actual positives), with BiodivBERT achieving an F1-score of 77.16% for the NER task, while BERT excelled in the RE task with an F1-score of 81.28%. Rigorous hyperparameter optimization further enhanced the performance of BiodivBERT in the RE task by 3.38%. The BiodivNERE corpora by Abdelmageed et al. (2022) were used to fine-tune several models for NER and RE tasks in the biodiversity domain. The first corpus from the BiodivNERE corpora is BiodivNER, which is a gold standard dataset (manually labelled test corpora) for evaluating NER tasks. The fine-tuning process employed the token classification method from the Hugging Face library (Hugging Face 2023b), which assigns labels to each token in a sequence. Experiments were conducted with a batch size of four, meaning the model processes four examples/rows of data at a time before making an update to improve its learning. This is due to the constraints of the NVIDIA® GeForce RTX™ 3060 graphics processor. (NVIDIA 2024) Model performance was evaluated using the seqeval library (Nakayama 2018), focusing on accuracy, precision, recall, and F1 scores. For text classification, the second corpus, BiodivRE, was utilized, following previous research recommendations to explore fine-tuning settings for BiodivBERT. Hyperparameter optimization (Feurer and Hutter 2019) was conducted using Hugging Face’s Trainer API with an Optuna backend (Hugging Face 2023a), concentrating on learning rate and the number of training epochs (i.e., the number of complete passes through the entire dataset during model training). The BiodiViz Knowledge Graph Application The fine-tuned NER and RE models with the best F1-scores—BiodivBERT and BERT, respectively—were integrated into the knowledge graph application. Fig. 1 illustrates the flowchart of the application pipeline. Each sentence in the input text will go through the NER model to identify and label the entities within the sentence. Subsequently, these labeled entities, together with the original sentence, will be input into the RE model. The RE model will analyze every pair of entities for a potential relation and output the type of relation they share. The application will then utilize this data to create a graph with appropriate labels and color-coding. An example of the application's user interface with the knowledge graph is shown in Fig. 2. This study highlights the practical application of machine learning and data visualization in advancing biodiversity research, emphasizing the importance of developing user-friendly tools to support scientific exploration and discovery. The BiodiViz application, including the code and resources, is available on GitHub*1, providing an accessible tool for biodiversity researchers to streamline their analyses.

Arabic word tokenization system using the maximum matching model

Word tokenization is the first stage for higher-order Natural Language Processing (NLP) tasks like Part-of-Speech (PoS) tagging, parsing, and named entity recognition. The amount of text on the World Wide Web is growing daily in the present era of technology, necessitating the use of advanced instruments. Since more and more people speak Arabic around the world, Arabic language processing systems must be improved. Due to the writing style of Arabic with a lack of support for capitalization features and the use of compound words, it is difficult to perform word tokenization. This research paper proposes a novel Arabic word tokenization system based on the knowledge. To develop this system, a maximum matching model with its two variations, namely forward and reverse maximum matching is used. The proposed system is implemented in Python. The results produced during system evaluation report high performance.

Pictorial Review of Rare Pancreatic Tumors and Tumor-like Lesions: Radiologic–Pathologic Correlation

Rare pancreatic tumors and non-neoplastic tumor-like lesions present a diagnostic challenge due to their uncommon occurrence and overlapping imaging characteristics with more prevalent pancreatic neoplasms. Advances in imaging technologies and diagnostic criteria have contributed to increased detection of these rare entities in clinical practice. This pictorial review focuses on the radiologic–pathologic correlation of rare pancreatic tumors, including colloid carcinoma, acinar cell carcinoma, pancreatoblastoma, primary pancreatic lymphoma, and non-neoplastic tumor-like lesions such as hamartomas and inflammatory pseudotumors. Detailed imaging features, such as signal intensities on MRI and enhancement patterns on CT, are correlated with pathological findings to assist in the differential diagnosis. Familiarity with these characteristics is crucial for radiologists to ensure accurate diagnosis and guide appropriate treatment strategies, as management and prognosis significantly differ from common pancreatic neoplasms.

Dual-tiered insights: cross-examining entities in free text electronic health records

Abstract Background Poor determination and processing of actions at the primary care level are found to be associated with unstructured and cluttered electronic health records (EHRs). Physicians recommend that EHRs must highlight entities, such as action items, findings, follow-ups, and reasons for medication changes. There’s consensus around the success of large language models (LLMs) in detecting clinical and non-clinical entities in EHRs. However, end-to-end systems make limited efforts to go beyond entity and relationship detection to support clinical decision-making. Methods We propose a two-tier system that generates case-based comparative reports for entities highlighted by physicians in EHRs. First, we fine-tuned a custom Bidirectional Encoder Representations from Transformers (BERT) model that detects the required entities within the EHRs. Second, a clustering-based service generates a report comparing the detected entities of a specific EHR with other EHRs sharing the same conditions determined by the healthcare professional. Results The system leverages the MIMIC IV deidentified free-text clinical notes dataset. The first-tier successfully classified entities with a macro-score average range of [0.87 - 0.94]. For the second tier, preliminary simulations on a set of test samples were made to cluster the entities. The simulations leverage a custom statistical tests engine offering a suite of entity-type specific tests. The results of these tests are used to generate the case-based detected entities comparative report. Conclusions Highlighting clinical and non-clinical entities bridges the communication gap between primary and secondary care. Clustering these entities extends the capability of LLMs in clinical support decision systems by providing healthcare professionals with the ability to run comparative analyses between EHRs. The comparative analysis reports provide a factual comparison between a selected number of detected entities and their thresholds. Key messages • Clustering clinical and non-clinical entities extends the capacity of named entity recognition in clinical decision-making systems. • Ability to generate a comparative analysis report based on figures extracted from recognised entities in free-text electronic health records.

Socioeconomic Status and Local Biodiversity as Predictors of Attitudes towards Nature among High School Students in a Latin American Metropolis.

Urbanization and biodiversity loss disconnect humans from nature and may limit pro-environmental actions. Although, a growing body of knowledge exists on people's attitudes towards nature, the majority focuses on the influence of socio-environmental factors. Significant variations in socioeconomic and educational levels in urban areas affect access to biodiversity. In this study, we aim to identify the socio-environmental variables that are associated with people's attitudes towards nature (student's knowledge about urban birds and their participation in conservation actions). We collected data on local biodiversity, socioeconomical status, and high school students' attitudes towards nature (15 schools, n = 175 students) in a large Latin American capital city. Structural equation modeling (SEM) was utilized to identify relationships among all variables. Socioeconomic characterization was assessed using levels of socio-materiality and school performance scores. Definitions of ecological complexity were based on bird species richness and vegetation density. Knowledge of urban birds was evaluated using two indicators: urban bird identification (recognition of names) and identification of the biogeographic origin of urban birds (native or exotic). Participation in conservation actions by students was assessed based on implementation scale and implementation frequency. According to our results, respondents' knowledge of birds was poor: 42% of students identified only two species (the domestic pigeon and the monk parakeet, both exotic). Only 26% of total students participated in conservation actions. SEM analysis revealed that socioeconomic status was positively associated with the ecological complexity of surrounding areas. In turn, ecological complexity showed a positive relationship with students' knowledge on urban birds. Finally, students' knowledge of urban birds showed a positive influence on conservation actions. Our findings suggest that, the improvement of students' knowledge about local biodiversity and ensuring equitable access to biodiversity could be an effective way to foster students' involvement in conservation actions, especially in neighborhoods of lower socioeconomic levels.

Syntax and the brain: language evolution as the missing link(ing theory)?

IntroductionThis paper provides proof of concept that neurolinguistic research on human language syntax would benefit greatly by expanding its scope to include evolutionary considerations, as well as non-propositional functions of language, including naming/nicknaming and verbal aggression. In particular, an evolutionary approach can help circumvent the so-called granularity problem in studying the processing of syntax in the brain, that is, the apparent mismatch between the abstract postulates of syntax (e.g. Tense Phrase (TP), Determiner Phrase (DP), etc.) and the concrete units of neurobiology (neurons, axons, etc.).MethodsFirst, we decompose syntax into its evolutionary primitives, identifying one of the earliest stages as a simple, flat combination of just one verb and one noun. Next, we identify proxies (“living fossils”) of such a stage in present-day languages, including compounds and small clauses, lacking at least some layers of structure, e.g. TPs and DPs. These proxies of ancestral language have been subjected to fMRI neuroimaging experiments.ResultsWe discuss the finding that less hierarchical small clauses, in contrast to full sentences with TPs and DPs, show reduced activation in the left Broca’s area (BA) 44 and the right basal ganglia, consistent with the hypothesis that more recent, more elaborate syntax requires more connectivity in the Broca’s-basal ganglia network, whose neuronal density has been significantly enhanced in recent evolution, implicating mutations in FOXP2 and other genes. We also discuss the finding that the processing of ancestral verb-noun compounds, which are typically used for (derogatory) naming and nicknaming, shows enhanced activation in the right fusiform gyrus area (BA 37), the area that is implicated in the processing of metaphoricity and imageability, but also in naming and face recognition, opening up an intriguing possibility that the enhanced face recognition in humans was facilitated by the early emergence of a simple syntactic strategy for naming.DiscussionThe considerations in this paper are consistent with the hypothesis of a gradual gene-culture co-evolution of syntax and the brain, targeting cortico-striatal brain networks. It is also of note that a sound grounding in neurobiology of language should in turn inform syntactic theories themselves.

Accelerating Evidence Synthesis in Observational Studies: Development of a Living Natural Language Processing-Assisted Intelligent Systematic Literature Review System.

Systematic literature review (SLR), a robust method to identify and summarize evidence from published sources, is considered to be a complex, time-consuming, labor-intensive, and expensive task. This study aimed to present a solution based on natural language processing (NLP) that accelerates and streamlines the SLR process for observational studies using real-world data. We followed an agile software development and iterative software engineering methodology to build a customized intelligent end-to-end living NLP-assisted solution for observational SLR tasks. Multiple machine learning-based NLP algorithms were adopted to automate article screening and data element extraction processes. The NLP prediction results can be further reviewed and verified by domain experts, following the human-in-the-loop design. The system integrates explainable articificial intelligence to provide evidence for NLP algorithms and add transparency to extracted literature data elements. The system was developed based on 3 existing SLR projects of observational studies, including the epidemiology studies of human papillomavirus-associated diseases, the disease burden of pneumococcal diseases, and cost-effectiveness studies on pneumococcal vaccines. Our Intelligent SLR Platform covers major SLR steps, including study protocol setting, literature retrieval, abstract screening, full-text screening, data element extraction from full-text articles, results summary, and data visualization. The NLP algorithms achieved accuracy scores of 0.86-0.90 on article screening tasks (framed as text classification tasks) and macroaverage F1 scores of 0.57-0.89 on data element extraction tasks (framed as named entity recognition tasks). Cutting-edge NLP algorithms expedite SLR for observational studies, thus allowing scientists to have more time to focus on the quality of data and the synthesis of evidence in observational studies. Aligning the living SLR concept, the system has the potential to update literature data and enable scientists to easily stay current with the literature related to observational studies prospectively and continuously.

Chronobridge: a novel framework for enhanced temporal and relational reasoning in temporal knowledge graphs

The task of predicting entities and relations in Temporal Knowledge Graph (TKG) extrapolation is crucial and has been studied extensively. Mainstream algorithms, such as Gated Recurrent Unit (GRU) models, primarily focus on encoding historical factual features within TKGs, often neglecting the importance of incorporating entities and relational features during decoding. This bias ultimately leads to loss of detail and inadequate prediction accuracy during the inference process. To address this issue, a novel ChronoBridge framework is proposed that features a dual mechanism of a chronological node encoder and a bridged feature fusion decoder. Specifically, the chronological node encoder employs an advanced recursive neural network with an enhanced GRU in an autoregressive manner to model historical KG sequences, thereby accurately capturing entity changes over time and significantly enhancing the model’s ability to identify and encode temporal patterns of facts across the timeline. Meanwhile, the bridged feature fusion decoder utilizes a new variant of GRU and a multilayer perception mechanism during the prediction phase to extract entity and relation features and fuse them for inference, thereby strengthening the reasoning capabilities of the model for future events. Testing on three standard datasets showed significant improvements, with a 25.21% increase in MRR accuracy and a 39.38% enhancement in relation inference. This advancement not only improves the understanding of temporal evolution in knowledge graphs but also sets a foundation for future research and applications of TKG reasoning.

Health Care Language Models and Their Fine-Tuning for Information Extraction: Scoping Review.

In response to the intricate language, specialized terminology outside everyday life, and the frequent presence of abbreviations and acronyms inherent in health care text data, domain adaptation techniques have emerged as crucial to transformer-based models. This refinement in the knowledge of the language models (LMs) allows for a better understanding of the medical textual data, which results in an improvement in medical downstream tasks, such as information extraction (IE). We have identified a gap in the literature regarding health care LMs. Therefore, this study presents a scoping literature review investigating domain adaptation methods for transformers in health care, differentiating between English and non-English languages, focusing on Portuguese. Most specifically, we investigated the development of health care LMs, with the aim of comparing Portuguese with other more developed languages to guide the path of a non-English-language with fewer resources. This study aimed to research health care IE models, regardless of language, to understand the efficacy of transformers and what are the medical entities most commonly extracted. This scoping review was conducted using the PRISMA-ScR (Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews) methodology on Scopus and Web of Science Core Collection databases. Only studies that mentioned the creation of health care LMs or health care IE models were included, while large language models (LLMs) were excluded. The latest were not included since we wanted to research LMs and not LLMs, which are architecturally different and have distinct purposes. Our search query retrieved 137 studies, 60 of which met the inclusion criteria, and none of them were systematic literature reviews. English and Chinese are the languages with the most health care LMs developed. These languages already have disease-specific LMs, while others only have general-health care LMs. European Portuguese does not have any public health care LM and should take examples from other languages to develop, first, general-health care LMs and then, in an advanced phase, disease-specific LMs. Regarding IE models, transformers were the most commonly used method, and named entity recognition was the most popular topic, with only a few studies mentioning Assertion Status or addressing medical lexical problems. The most extracted entities were diagnosis, posology, and symptoms. The findings indicate that domain adaptation is beneficial, achieving better results in downstream tasks. Our analysis allowed us to understand that the use of transformers is more developed for the English and Chinese languages. European Portuguese lacks relevant studies and should draw examples from other non-English languages to develop these models and drive progress in AI. Health care professionals could benefit from highlighting medically relevant information and optimizing the reading of the textual data, or this information could be used to create patient medical timelines, allowing for profiling.

An editorial note on extraction and evaluation of knowledge entities from scientific documents

An editorial note on extraction and evaluation of knowledge entities from scientific documents

Sentence Embeddings and Semantic Entity Extraction for Identification of Topics of Short Fact-Checked Claims

The objective of this research was to design a method to assign topics to claims debunked by fact-checking agencies. During the fact-checking process, access to more structured knowledge is necessary; therefore, we aim to describe topics with semantic vocabulary. Classification of topics should go beyond simple connotations like instance-class and rather reflect broader phenomena that are recognized by fact checkers. The assignment of semantic entities is also crucial for the automatic verification of facts using the underlying knowledge graphs. Our method is based on sentence embeddings, various clustering methods (HDBSCAN, UMAP, K-means), semantic entity matching, and terms importance assessment based on TF-IDF. We represent our topics in semantic space using Wikidata Q-ids, DBpedia, Wikipedia topics, YAGO, and other relevant ontologies. Such an approach based on semantic entities also supports hierarchical navigation within topics. For evaluation, we compare topic modeling results with claims already tagged by fact checkers. The work presented in this paper is useful for researchers and practitioners interested in semantic topic modeling of fake news narratives.

Dual Contrastive Learning for Cross-Domain Named Entity Recognition

Benefiting many information retrieval applications, named entity recognition (NER) has shown impressive progress. Recently, there has been a growing trend to decompose complex NER tasks into two subtasks (e.g., entity span detection (ESD) and entity type classification (ETC), to achieve better performance. Despite the remarkable success, from the perspective of representation, existing methods do not explicitly distinguish non-entities and entities, which may lead to ESD errors. Meanwhile, they do not explicitly distinguish entities with different entity types, which may lead to entity type misclassification. As such, the limited representation abilities may challenge some competitive NER methods, leading to unsatisfactory performance, especially in the low-resource setting (e.g., cross-domain NER). In light of these challenges, we propose to utilize contrastive learning to refine the original chaotic representations and learn the generalized representations for cross-domain NER. In particular, this article proposes a dual contrastive learning model (Dual-CL), which respectively utilizes a token-level contrastive learning module and a sentence-level contrastive learning module to enhance ESD, ETC for cross-domain NER. Empirical results on 10 domain pairs under two different settings show that Dual-CL achieves better performances than compared baselines in terms of several standard metrics. Moreover, we conduct detailed analyses to are presented to better understand each component’s effectiveness.

Sequential lexicon enhanced bidirectional encoder representations from transformers: Chinese named entity recognition using sequential lexicon enhanced BERT

Lexicon Enhanced Bidirectional Encoder Representations from Transformers (LEBERT) has achieved great success in Chinese Named Entity Recognition (NER). LEBERT performs lexical enhancement with a Lexicon Adapter layer, which facilitates deep lexicon knowledge fusion at the lower layers of BERT. However, this method is likely to introduce noise words and does not consider the possible conflicts between words when fusing lexicon information. To address this issue, we advocate for a novel lexical enhancement method, Sequential Lexicon Enhanced BERT (SLEBERT) for the Chinese NER, which builds sequential lexicon to reduce noise words and resolve the problem of lexical conflict. Compared with LEBERT, it leverages the position encoding of sequential lexicon and adaptive attention mechanism of sequential lexicon to enhance the lexicon feature. Experiments on the four available datasets identified that SLEBERT outperforms other lexical enhancement models in performance and efficiency.

Design of agricultural question answering information extraction method based on improved BILSTM algorithm

With the rapid growth of the agricultural information and the need for data analysis, how to accurately extract useful information from massive data has become an urgent first step in agricultural data mining and application. In this study, an agricultural question-answering information extraction method based on the BE-BILSTM (Improved Bidirectional Long Short-Term Memory) algorithm is designed. Firstly, it uses Python’s Scrapy crawler framework to obtain the information of soil types, crop diseases and pests, and agricultural trade information, and remove abnormal values. Secondly, the information extraction converts the semi-structured data by using entity extraction methods. Thirdly, the BERT (Bidirectional Encoder Representations from Transformers) algorithm is introduced to improve the performance of the BILSTM algorithm. After comparing with the BERT-CRF (Conditional Random Field) and BILSTM algorithm, the result shows that the BE-BILSTM algorithm has better information extraction performance than the other two algorithms. This study improves the accuracy of the agricultural information recommendation system from the perspective of information extraction. Compared with other work that is done from the perspective of recommendation algorithm optimization, it is more innovative; it helps to understand the semantics and contextual relationships in agricultural question and answer, which improves the accuracy of agricultural information recommendation systems. By gaining a deeper understanding of farmers’ needs and interests, the system can better recommend relevant and practical information.

Evaluating Medical Entity Recognition in Health Care: Entity Model Quantitative Study.

Named entity recognition (NER) models are essential for extracting structured information from unstructured medical texts by identifying entities such as diseases, treatments, and conditions, enhancing clinical decision-making and research. Innovations in machine learning, particularly those involving Bidirectional Encoder Representations From Transformers (BERT)-based deep learning and large language models, have significantly advanced NER capabilities. However, their performance varies across medical datasets due to the complexity and diversity of medical terminology. Previous studies have often focused on overall performance, neglecting specific challenges in medical contexts and the impact of macrofactors like lexical composition on prediction accuracy. These gaps hinder the development of optimized NER models for medical applications. This study aims to meticulously evaluate the performance of various NER models in the context of medical text analysis, focusing on how complex medical terminology affects entity recognition accuracy. Additionally, we explored the influence of macrofactors on model performance, seeking to provide insights for refining NER models and enhancing their reliability for medical applications. This study comprehensively evaluated 7 NER models-hidden Markov models, conditional random fields, BERT for Biomedical Text Mining, Big Transformer Models for Efficient Long-Sequence Attention, Decoding-enhanced BERT with Disentangled Attention, Robustly Optimized BERT Pretraining Approach, and Gemma-across 3 medical datasets: Revised Joint Workshop on Natural Language Processing in Biomedicine and its Applications (JNLPBA), BioCreative V CDR, and Anatomical Entity Mention (AnatEM). The evaluation focused on prediction accuracy, resource use (eg, central processing unit and graphics processing unit use), and the impact of fine-tuning hyperparameters. The macrofactors affecting model performance were also screened using the multilevel factor elimination algorithm. The fine-tuned BERT for Biomedical Text Mining, with balanced resource use, generally achieved the highest prediction accuracy across the Revised JNLPBA and AnatEM datasets, with microaverage (AVG_MICRO) scores of 0.932 and 0.8494, respectively, highlighting its superior proficiency in identifying medical entities. Gemma, fine-tuned using the low-rank adaptation technique, achieved the highest accuracy on the BioCreative V CDR dataset with an AVG_MICRO score of 0.9962 but exhibited variability across the other datasets (AVG_MICRO scores of 0.9088 on the Revised JNLPBA and 0.8029 on AnatEM), indicating a need for further optimization. In addition, our analysis revealed that 2 macrofactors, entity phrase length and the number of entity words in each entity phrase, significantly influenced model performance. This study highlights the essential role of NER models in medical informatics, emphasizing the imperative for model optimization via precise data targeting and fine-tuning. The insights from this study will notably improve clinical decision-making and facilitate the creation of more sophisticated and effective medical NER models.

Exploring the Capabilities of ChatGPT in Ancient Chinese Translation and Person Name Recognition

Abstract This study examines the potential of ChatGPT, notable for its skill in standard modern languages, in the context of ancient Chinese. We focus on two fundamental tasks: translating ancient Chinese into modern Chinese and identifying ancient Chinese names. We conduct experiments on Shi Shuo Xin Yu, which is a collection of anecdotes, short conversations, and pithy observations on personalities who lived in China between about 150 and 420 A.D. We also include other methods, like Baidu Yiyan (ERNIE-Bot) for ancient-to-modern translation, and SIKU-RoBERTa for ancient Chinese name recognition. The results reveal that ChatGPT’s performance in ancient Chinese remains in need of enhancement, but it excels in the task of ancient-to-modern translation with the provision of a few context sentences. For the ancient Chinese name recognition, ChatGPT can outperform SIKU-RoBERTa model in terms of precision, but suffers from low recall.

Lifestyle factors in the biomedical literature: An ontology and comprehensive resources for named entity recognition.

Despite lifestyle factors (LSFs) being increasingly acknowledged in shaping individual health trajectories, particularly in chronic diseases, they have still not been systematically described in the biomedical literature. This is in part because no named entity recognition (NER) system exists, which can comprehensively detect all types of LSFs in text. The task is challenging due to their inherent diversity, lack of a comprehensive LSF classification for dictionary-based NER, and lack of a corpus for deep learning-based NER. We present a novel Lifestyle Factor Ontology (LSFO), which we used to develop a dictionary-based system for recognition and normalization of LSFs. Additionally, we introduce a manually annotated corpus for LSFs (LSF200) suitable for training and evaluation of NER systems, and use it to train a transformer-based system. Evaluating the performance of both NER systems on the corpus revealed an F-score of 64% for the dictionary-based system and 76% for the transformer-based system. Large-scale application of these systems on PubMed abstracts and PMC Open Access articles identified over 300 million mentions of LSF in the biomedical literature. LSFO, the annotated LSF200 corpus, and the detected LSFs in PubMed and PMC-OA articles using both NER systems, are available under open licenses via the following GitHub repository: Https://github.com/EsmaeilNourani/LSFO-expansion. This repository contains links to two associated GitHub repositories and a Zenodo project related to the study. LSFO is also available at BioPortal: Https://bioportal.bioontology.org/ontologies/LSFO. Supplementary data are available at Bioinformatics online.