Recognition Step Research Articles

CRISPR-Cas12a bends DNA to destabilize base pairs during target interrogation.

RNA-guided endonucleases are involved in processes ranging from adaptive immunity to site-specific transposition and have revolutionized genome editing. CRISPR-Cas9, -Cas12 and related proteins use guide RNAs to recognize ∼20-nucleotide target sites within genomic DNA by mechanisms that are not yet fully understood. We used structural and biochemical methods to assess early steps in DNA recognition by Cas12a protein-guide RNA complexes. We show here that Cas12a initiates DNA target recognition by bending DNA to induce transient nucleotide flipping that exposes nucleobases for DNA-RNA hybridization. Cryo-EM structural analysis of a trapped Cas12a-RNA-DNA surveillance complex and fluorescence-based conformational probing show that Cas12a-induced DNA helix destabilization enables target discovery and engagement. This mechanism of initial DNA interrogation resembles that of CRISPR-Cas9 despite distinct evolutionary origins and different RNA-DNA hybridization directionality of these enzyme families. Our findings support a model in which RNA-mediated DNA interference begins with local helix distortion by transient CRISPR-Cas protein binding.

SegmOnto: A Controlled Vocabulary to Describe and Process Digital Facsimiles

Our initiative aims at designing a controlled vocabulary for the description of the layout of textual sources: SegmOnto. Following a more physical approach rather than a strictly semantic one, it is designed as a pragmatic and generic typology, coping with most of the Western historical documents rather than answering specific needs. The harmonisation of the layout description has a double objective: on the one hand it facilitates the mutualisation of annotated data and therefore the training of better models for page segmentation (a crucial preliminary step for text recognition), on the other hand it allows the development of a shared post-processing workflow and pipeline for the transformation of ALTO or PAGE files into DH standard formats, which preserves as much as possible the link between the extracted information and the digital facsimile. To demonstrate the capacity of SegmOnto to answer both these objectives, we aggregate data from multiple projects to train a layout analysis model, and we propose a prototype of a generic pipeline for converting ALTO-XMLs into XML-TEI.

Further Evidence That Female Anoplophora glabripennis (Coleoptera: Cerambycidae) Utilizes Photo-Degradation to Produce Volatiles That Are Attractive to Adult Males

The Asian longhorned beetle, Anoplophora glabripennis (Coleoptera: Cerambycidae: Lamiinae), is a serious pest of over 43 species of hardwood trees in North America, China and Europe. The development of an effective lure and trap for monitoring A. glabripennis has been hindered by the fact that mate finding involves a rather complex series of behaviors and responses to several chemical (and visual), cues. Adults (female-biased) locate a tree via host kairomones. Research has demonstrated that female contact pheromone components are precursors that undergo abiotic oxidation to yield attractive volatile components. Males also produce a pheromone to attract other adults nearby before the final step of recognition by males to the female-produced trail and contact pheromones. Our research aimed to identify new female-produced components from ozone and UV-treated A. glabripennis body washes and test them for behavioral activity using laboratory and field assays. The ozone and UV treatment of virgin female extract yielded sixteen aldehydes, nine of which were found in trace amounts. All sixteen aldehydes elicited antennal responses (GC-EAD) in both males and females, although responses were clearer and more distinct with standards of hexanal, heptanal, octanal, nonanal, decanal, undecanal and dodecanal. Olfactometer assays showed that males were highly attracted to a blend of these seven aldehydes. Females did not show any attraction to the blend in olfactometer assays. Despite low population levels during field tests in 2018 in China, traps containing the aldehyde blend detected A. glabripennis on a weekly basis over 6 weeks and caught significantly more adults (mainly males) than control ‘flight intercept panel’ traps. Field tests in China in 2019 that used the aldehyde blend along with a three-component host blend lure also caught significantly more males (nearly 5×) than blank control traps.

Simulation of body trajectory detection in sports training based on infrared thermal imaging image features

The traditional motion trajectory detection methods are limited by illumination conditions and occlusion, and it is difficult to capture the athlete’s motion path accurately and in real time. The aim of this study is to improve the accuracy and real-time performance of motion trajectory detection, and to provide more effective auxiliary means for athletes’ training and competition. The image feature analysis method based on Infrared Thermal Imaging, combined with computer vision and machine learning technology, is used to monitor and analyze the motion scene. Infrared Thermal Imaging equipment is used to collect data of motion scenes and obtain infrared image sequence. Then, through the steps of image preprocessing, feature extraction and target recognition, the position and motion trajectory of athletes at different time points are obtained. The experimental results show that the image feature analysis method based on Infrared Thermal Imaging can effectively capture the athlete’s motion trajectory, and has higher accuracy and real-time performance. The method has better adaptability and robustness in complex environments such as lighting conditions and occlusion.

Structural and mechanistic insights into a mesophilic prokaryotic Argonaute.

Argonaute (Ago) proteins are programmable nucleases found in all domains of life, playing a crucial role in biological processes like DNA/RNA interference and gene regulation. Mesophilic prokaryotic Agos (pAgos) have gained increasing research interest due to their broad range of potential applications, yet their molecular mechanisms remain poorly understood. Here, we present seven cryo-electron microscopy structures of Kurthia massiliensis Ago (KmAgo) in various states. These structures encompass the steps of apo-form, guide binding, target recognition, cleavage, and release, revealing that KmAgo employs a unique DDD catalytic triad, instead of a DEDD tetrad, for DNA target cleavage under 5'P-DNA guide conditions. Notably, the last catalytic residue, D713, is positioned outside the catalytic pocket in the absence of guide. After guide binding, D713 enters the catalytic pocket. In contrast, the corresponding catalytic residue in other Agos has been consistently located in the catalytic pocket. Moreover, we identified several sites exhibiting enhanced catalytic activity through alanine mutagenesis. These sites have the potential to serve as engineering targets for augmenting the catalytic efficiency of KmAgo. This structural analysis of KmAgo advances the understanding of the diversity of molecular mechanisms by Agos, offering insights for developing and optimizing mesophilic pAgos-based programmable DNA and RNA manipulation tools.

Deep handwritten diagram segmentation

AbstractHandwriting is a natural way to communicate and exchange ideas, but converting handwritten diagrams to application‐specific digital formats requires skill and time. Automatic handwritten document conversion can save time, but diagrams and text require different recognition engines. Since accurate segmentation of handwritten diagrams can improve the accuracy of later diagram recognition steps, the authors propose to solve the problem of segmentation of text and non‐text elements of handwritten diagrams using deep semantic segmentation. The model, DeepDP is a flexible U‐net style architecture that can be tuned in complexity to a level appropriate for a particular dataset and diagram type. Experiments on a public hand‐drawn flowchart dataset and a business process diagram dataset show excellent performance, with a per pixel accuracy of 98.6% on the public flowchart datasets and improvement over the 99.3% text stroke accuracy and 96.6% non‐text stroke accuracy obtained by state of the art methods that use online stroke information. On the smaller offline business process diagram dataset, the method obtains a per‐pixel accuracy of 96.9%.

Auto-segmentation of hemi-diaphragms in free-breathing dynamic MRI of pediatric subjects with thoracic insufficiency syndrome.

In respiratory disorders such as thoracic insufficiency syndrome (TIS), the quantitative study of the regional motion of the left hemi-diaphragm (LHD) and right hemi-diaphragm (RHD) can give detailed insights into the distribution and severity of the abnormalities in individual patients. Dynamic magnetic resonance imaging (dMRI) is a preferred imaging modality for capturing dynamic images of respiration since dMRI does not involve ionizing radiation and can be obtained under free-breathing conditions. Using 4D images constructed from dMRI of sagittal locations, diaphragm segmentation is an evident step for the said quantitative analysis of LHD and RHD in these 4D images. In this paper, we segment the LHD and RHD in three steps: recognition of diaphragm, delineation of diaphragm, and separation of diaphragm along the mid-sagittal plane into LHD and RHD. The challenges involved in dMRI images are low resolution, motion blur, suboptimal contrast resolution, inconsistent meaning of gray-level intensities for the same object across multiple scans, and low signal-to-noise ratio. We have utilized deep learning (DL) concepts such as Path Aggregation Network and Dual Attention Network for the recognition step, Dense-Net and Residual-Net in an enhanced encoder-decoder architecture for the delineation step, and a combination of GoogleNet and Recurrent Neural Network for the identification of the mid-sagittal plane in the separation step. Due to the challenging images of TIS patients attributed to their highly distorted and variable anatomy of the thorax, in such images we localize the diaphragm using the auto-segmentations of the lungs and the thoraco-abdominal skin. We achieved an average±SD mean-Hausdorff distance of ∼3±3 mm for the delineation step and a positional error of ∼3±3 mm in recognizing the mid-sagittal plane in 100 3D test images of TIS patients with a different set of ∼430 3D images of TIS patients utilized for building the models for delineation, and separation. We showed that auto-segmentations of the diaphragm are indistinguishable from segmentations by experts, in images of near-normal subjects. In addition, the algorithmic identification of the mid-sagittal plane is indistinguishable from its identification by experts in images of near-normal subjects. Motivated by applications in surgical planning for disorders such as TIS, we have shown an auto-segmentation set-up for the diaphragm in dMRI images of TIS pediatric subjects. The results are promising, showing that our system can handle the aforesaid challenges. We intend to use the auto-segmentations of the diaphragm to create the initial ground truth (GT) for newly acquired data and then refining them, to expedite the process of creating GT for diaphragm motion analysis, and to test the efficacy of our proposed method to optimize pre-treatment planning and post-operative assessment of patients with TIS and other disorders.

Relative genotoxicity of polycyclic aromatic hydrocarbons inferred from free energy perturbation approaches

Utilizing molecular dynamics and free energy perturbation, we examine the relative binding affinity of several covalent polycyclic aromatic hydrocarbon - DNA (PAH-DNA) adducts at the central adenine of NRAS codon-61, a mutational hotspot implicated in cancer risk. Several PAHs classified by the International Agency for Research on Cancer as probable, possible, or unclassifiable as to carcinogenicity are found to have greater binding affinity than the known carcinogen, benzo[a]pyrene (B[a]P). van der Waals interactions between the intercalated PAH and neighboring nucleobases, and minimal disruption of the DNA duplex drive increases in binding affinity. PAH-DNA adducts may be repaired by global genomic nucleotide excision repair (GG-NER), hence we also compute relative free energies of complexation of PAH-DNA adducts with RAD4-RAD23 (the yeast ortholog of human XPC-RAD23) which constitutes the recognition step in GG-NER. PAH-DNA adducts exhibiting the greatest DNA binding affinity also exhibit the least RAD4-RAD23 complexation affinity and are thus predicted to resist the GG-NER machinery, contributing to their genotoxic potential. In particular, the fjord region PAHs dibenzo[a,l]pyrene, benzo[g]chrysene, and benzo[c]phenanthrene are found to have greater binding affinity while having weaker RAD4-RAD23 complexation affinity than their respective bay region analogs B[a]P, chrysene, and phenanthrene. We also find that the bay region PAHs dibenzo[a,j]anthracene, dibenzo[a,c]anthracene, and dibenzo[a,h]anthracene exhibit greater binding affinity and weaker RAD4-RAD23 complexation affinity than B[a]P. Thus, the study of PAH genotoxicity likely needs to be substantially broadened, with implications for public policy and the health sciences. This approach can be broadly applied to assess factors contributing to the genotoxicity of other unclassified compounds.

Application of Artificial Neural Networks to a Model of a Helicopter Rotor Blade for Damage Identification in Realistic Load Conditions.

Monitoring the integrity of aeronautical structures is fundamental for safety. Structural Health Monitoring Systems (SHMSs) perform real-time monitoring functions, but their performance must be carefully assessed. This is typically done by introducing artificial damages to the components; however, such a procedure requires the production and testing of a large number of structural elements. In this work, the damage detection performance of a strain-based SHMS was evaluated on a composite helicopter rotor blade root, exploiting a Finite Element (FE) model of the component. The SHMS monitored the bonding between the central core and the surrounding antitorsional layer. A damage detection algorithm was trained through FE analyses. The effects of the load's variability and of the damage were decoupled by including a load recognition step in the algorithm, which was accomplished either with an Artificial Neural Network (ANN) or a calibration matrix. Anomaly detection, damage assessment, and localization were performed by using an ANN. The results showed a higher load identification and anomaly detection accuracy using an ANN for the load recognition, and the load set was recognized with a satisfactory accuracy, even in damaged blades. This case study was focused on a real-world subcomponent with complex geometrical features and realistic load conditions, which was not investigated in the literature and provided a promising approach to estimate the performance of a strain-based SHMS.

CRISPR-Cas12a bends DNA to destabilize base pairs during target interrogation.

RNA-guided endonucleases are involved in processes ranging from adaptive immunity to site-specific transposition and have revolutionized genome editing. CRISPR-Cas9, -Cas12 and related proteins use guide RNAs to recognize ~20-nucleotide target sites within genomic DNA by mechanisms that are not yet fully understood. We used structural and biochemical methods to assess early steps in DNA recognition by Cas12a protein-guide RNA complexes. We show here that Cas12a initiates DNA target recognition by bending DNA to induce transient nucleotide flipping that exposes nucleobases for DNA-RNA hybridization. Cryo-EM structural analysis of a trapped Cas12a-RNA-DNA surveillance complex and fluorescence-based conformational probing show that Cas12a-induced DNA helix destabilization enables target discovery and engagement. This mechanism of initial DNA interrogation resembles that of CRISPR-Cas9 despite distinct evolutionary origins and different RNA-DNA hybridization directionality of these enzyme families. Our findings support a model in which RNA-mediated DNA engineering begins with local helix distortion by transient CRISPR-Cas protein binding.

Improving deep learning based bluespotted ribbontail ray (Taeniura Lymma) recognition

This paper presents the novel task of bluespotted ribbontail (BR) ray (Taeniura lymma) recognition using deep learning based computer vision methods to enable the identification of specific individuals of this species. Mapping the specific individuals in relation to location and time will allow marine researchers to understand their movement patterns, habitat choice, life span, size of the population and more – data which could allow monitoring and establishing a tailor-made conservation plan for this species.Our work is pioneer on this recognition problem. We give a detailed description of the three basic steps of detection, feature extraction and recognition in this vision problem and perform experiments to explore the system configuration and what improves the performance. A feature extraction enhancement as well as a crucial effect of a split into different main poses are demonstrated. Though the precision results achieved in this paper are still moderate and should be further improved, they are nevertheless promising and reasonable for practical use if the six best matches are chosen. For this scenario, almost 85% precision for upper-pose model, and almost 80% precision for left- and right-pose models, are achieved demonstrating the feasibility of the pipeline suggested as well as opportunities for improvement.

Study of the RNA splicing kinetics via in vivo 5-EU labeling.

Splicing is an important step of gene expression in all eukaryotes. Splice sites might be used with different efficiency, giving rise to alternative splicing products. At the same time, splice sites might be used at a variable rate. We used 5-ethynyl uridine labeling to sequence a nascent transcriptome of HeLa cells and deduced the rate of splicing for each donor and acceptor splice site. The following correlation analysis showed a correspondence of primary transcript features with the rate of splicing. Some dependencies we revealed were anticipated, such as a splicing rate decrease with a decreased complementarity of the donor splice site to U1 and acceptor sites to U2 snRNAs. Other dependencies were more surprising, like a negative influence of a distance to the 5' end on the rate of the acceptor splicing site utilization, or the differences in splicing rate between long, short, and RBM17-dependent introns. We also observed a deceleration of last intron splicing with an increase of the distance to the poly(A) site, which might be explained by the cooperativity of the splicing and polyadenylation. Additional analysis of splicing kinetics of SF3B4 knockdown cells suggested the impairment of a U2 snRNA recognition step. As a result, we deconvoluted the effects of several examined features on the splicing rate into a single regression model. The data obtained here are useful for further studies in the field, as they provide general splicing rate dependencies as well as help to justify the existence of slowly removed splice sites.

Construction of an Electrochemical Molecularly Imprinted Sensor Based on L-carnitine Doped Polypyrrole for Vibrio Parahaemolyticus Detection

This study addresses the pressing need for rapid detection of pathogenic microorganisms by designing and fabricating an L-carnitine-doped polypyrrole (L-CN/PPy) electrochemical molecularly imprinted sensor, with Vibrio parahaemolyticus as the model target. Employing L-carnitine as a dual-functional dopant and auxiliary recognition element in the electropolymerization process, the sensor not only enhances the conductive properties of polypyrrole but also significantly improves its specific recognition capability towards the target bacteria. The sensor preparation encompasses the electrochemical polymerization of L-CN/PPy composite material, immobilization and elution of template bacteria, followed by bacterial recognition and detection steps. Through systematic optimization of L-CN concentration, polymerization conditions, eluent composition, elution time, and operational parameters, the sensor's performance is maximized. Experimental results confirm the sensor's efficacy in rapidly completing template bacteria elution within just 30 minutes and achieving efficient recognition of Vibrio parahaemolyticus within 20 minutes, demonstrating high sensitivity and selectivity. Moreover, the straightforward and rapid construction method of the sensor introduces a novel technological strategy for practical microbial detection applications, highlighting extensive application prospects.

Construction of Financial Bill Recognition Model Based on Deep Learning

The common bills in life include VAT invoices, taxi invoices, train invoices, plane itineraries, machine-printed invoices, etc. Most of these common bills are presented in the form of fixed form templates, so template matching can be used. , for a certain fixed template bill, manually set the rules to determine the spatial position of the key area, extract the corresponding text information, or build a model with logical semantic relationship and spatial relative relationship between the bill texts of different attributes, from the global image of the image. Identify the required key text information in the text information. However, these methods are either limited by fixed ticket templates, or cannot guarantee considerable accuracy. The electronicization of paper invoices mainly needs to go through the steps of text detection, bill recognition and text recognition. Based on this, this paper adopts the DL method. Construct a financial bill recognition model and combine experiments to explore the effectiveness and superiority of the model. The results show that our model can achieve a recognition accuracy rate of up to 91\%, and also achieve a 26\% improvement in recognition speed.

Effectiveness of Mobile-Based Progressive and Fixed Physical Activity on Depression, Stress, Anxiety, and Quality of Life Outcomes Among Adults in South Korea: Randomized Controlled Trial.

Depression acts as a significant obstacle to the overall well-being of individuals. Given the significant consequences, timely recognition and proactive steps to manage symptoms of depression become essential. Such actions not only reduce personal distress but also play a crucial role in reducing its far-reaching impact on society as a whole. In response to this concern, the objective of this study was to explore the use of mobile-based interventions as a possible remedy. More specifically, this study aimed to investigate the effectiveness of 2 types of physical activity (PA), progressive and fixed, within a mobile-based app on depression, perceived stress, anxiety, physical health, and psychological health, aiming to contribute to the optimization of mental health benefits. Participants (N=60; mean age 25.29, SD 6.10 years) were recruited using a combination of web-based and offline methods, and the study lasted for 8 weeks. The baseline and posttest questionnaires were administered to all participants. The participants were randomly assigned to 1 of the 3 groups: progressive group (n=20; performing mobile-based progressive PA), fixed group (n=20; performing mobile-based fixed intensity PA), and control group C (n=20). Data analysis involved comparing scores between the experimental and control groups using a one-way ANOVA, paired sample t tests (2-tailed), and repeated measures ANOVA with a 3 (group)×2 (time) design. The findings revealed significant improvements in mental health indicators among participants engaged in both fixed and progressive PA groups compared with the control group. However, the fixed PA group demonstrated more significant reductions in symptoms. Specifically, the progressive PA group showed significant reductions in depression (F1,36=6.941; P=.01; ηp2=0.16) and perceived stress (F1,36=5.47; P=.03; ηp2=0.13), while the fixed PA group exhibited significant reductions in depression (F1,37=5.36; P=.03; ηp2=0.12), perceived stress (F1,37=7.81; P=.008; ηp2=0.17), and general anxiety disorder (F1,37=5.45; P=.03; ηp2=0.13) compared with the control group. This study underscores the potential of mobile-based PA in improving mental health outcomes. The findings offer significant insights for mental health professionals and researchers aiming to optimize mental well-being through innovative mobile therapies. Clinical Research Information Service KCT0009100; https://tinyurl.com/mr33fmur.

Design and analysis of face recognition system based on VGGFace-16 with various classifiers

This research presents a face recognition system based on different classifiers that deal with various face positions. The proposed system involves the extraction of features through the VGG-Face-16 deep neural network, which only extracts essential features of input images, leading to an improved recognition step and enhanced algorithm efficiency, while the recognition involves the radial basis function in support vector machine (SVM) classifier and evaluate the performance of the system. Also, the system is designed and implemented later by using other classifiers; they are K-neareste2 neighbour (KNN) classifiers, logistic regression (LR), gradient boosting (XGBoost), decision tree classifier (DT) and Naive Bayes classifier (NB). The proposed algorithm was tested with the four face databases: AT&T, PINs Face, linear friction welding (LFW) and real database. The database was divided into two groups: One contains a percentage of images that are used for training and the second contains a percentage of images (remainder) which was used for testing. The results show that the classification by RBF in SVM has the highest recognition rate in the case of using small, medium and large databases; it was 100% in AT&T and Real database, while its efficiency appears to be lower when using large-size databases whereas it is 96% in PINs database and 60.1% in LFW database.

The Recognition Pathway of the SARS-CoV-2 Spike Receptor-Binding Domain to Human Angiotensin-Converting Enzyme 2.

COVID-19 caused by SARS-CoV-2 has spread around the world. The receptor-binding domain (RBD) of the spike protein of SARS-CoV-2 is a critical component that directly interacts with host ACE2. Here, we simulate the ACE2 recognition processes of RBD of the WT, Delta, and OmicronBA.2 variants using our recently developed supervised Gaussian accelerated molecular dynamics (Su-GaMD) approach. We show that RBD recognizes ACE2 through three contact regions (regions I, II, and III), which aligns well with the anchor-locker mechanism. The higher binding free energy in State d of the RBDOmicronBA.2-ACE2 system correlates well with the increased infectivity of OmicronBA.2 in comparison with other variants. For RBDDelta, the T478K mutation affects the first step of recognition, while the L452R mutation, through its nearby Y449, affects the RBDDelta-ACE2 binding in the last step of recognition. For RBDOmicronBA.2, the E484A mutation affects the first step of recognition, the Q493R, N501Y, and Y505H mutations affect the binding free energy in the last step of recognition, mutations in the contact regions affect the recognition directly, and other mutations indirectly affect recognition through dynamic correlations with the contact regions. These results provide theoretical insights for RBD-ACE2 recognition and may facilitate drug design against SARS-CoV-2.

A Review of Document Binarization: Main Techniques, New Challenges, and Trends

Document image binarization is a challenging task, especially when it comes to text segmentation in degraded document images. The binarization, as a pre-processing step of Optical Character Recognition (OCR), is one of the most fundamental and commonly used segmentation methods. It separates the foreground text from the background of the document image to facilitate subsequent image processing. In view of the different degradation degrees of document images, researchers have proposed a variety of solutions. In this paper, we have summarized some challenges and difficulties in the field of document image binarization. Approximately 60 methods documenting image binarization techniques are mentioned, including traditional algorithms and deep learning-based algorithms. Here, we evaluated the performance of 25 image binarization techniques on the H-DIBCO2016 dataset to provide some help for future research.

Malayalam Handwritten Words Recognition: A Review

This review examines character segmentation and offers an elegant method for identifying and transforming handwritten Malayalam words from picture documents into text. Character touchings, different writing styles, and noisy, damaged scanned photos make it difficult to recognise handwritten text. Taking use of today's world of rich data and algorithmic developments, the system uses deep convolutional neural networks (CNNs) to address these challenges. The three steps of Malayalam handwritten word recognition are segmentation, recognition, and pre-processing. Making Malayalam character datasets is the first stage, and then pre-processing to improve image quality comes next. Then, in order to maximise the system's capacity to precisely forecast Malayalam characters, a CNN model is built to extract relevant information. The last phase of the recognition process involves the system classifying the characters. This project is significant since it uses CNN filters to enhance feature recognition, which enhances the accuracy of Malayalam character prediction. Key Words: Deep Learning, Deep Convolution Neural Network (DCNN), Character recognition, Character segmentation,

Impact of Translation on Biomedical Information Extraction: Experiment on Real-Life Clinical Notes.

Biomedical natural language processing tasks are best performed with English models, and translation tools have undergone major improvements. On the other hand, building annotated biomedical data sets remains a challenge. The aim of our study is to determine whether the use of English tools to extract and normalize French medical concepts based on translations provides comparable performance to that of French models trained on a set of annotated French clinical notes. We compared 2 methods: 1 involving French-language models and 1 involving English-language models. For the native French method, the named entity recognition and normalization steps were performed separately. For the translated English method, after the first translation step, we compared a 2-step method and a terminology-oriented method that performs extraction and normalization at the same time. We used French, English, and bilingual annotated data sets to evaluate all stages (named entity recognition, normalization, and translation) of our algorithms. The native French method outperformed the translated English method, with an overall F1-score of 0.51 (95% CI 0.47-0.55), compared with 0.39 (95% CI 0.34-0.44) and 0.38 (95% CI 0.36-0.40) for the 2 English methods tested. Despite recent improvements in translation models, there is a significant difference in performance between the 2 approaches in favor of the native French method, which is more effective on French medical texts, even with few annotated documents.