Hasil untuk "Computational linguistics. Natural language processing"

Isabel Cristina Alfonzo de Tovar

Jin Chencong, Cheng Defang, Bo Likang

Abstract Aiming at the defects of current traffic sign image recognition, such as high error rate and poor recognition in real time, the intelligent recognition method of traffic sign image based on visual communication technology is proposed with the main objective of improving the accuracy of traffic sign image recognition. This work proposes traffic sign recognition using visual communication preprocessing coupled with HOG and SVM-based classification, which attempts to overcome limitations in accuracy and in real time posed by present approaches. Firstly, the traffic sign images are collected and pre-processed according to the visual communication technique to improve the quality of traffic sign images; Then, based on the analysis of the traffic sign recognition system, a scheme for traffic sign detection and recognition is proposed. The system performs normalization operations on the captured images after greyscaling and smoothing, and unifies the image size to facilitate feature extraction. The pipeline filters noise using Gabor filtering, extracts features with HOG, and classifies features robustly with an SVM using an RBF kernel. Feature segmentation of the processed images and feature extraction using the HOG algorithm are performed. Finally, the SVM risk algorithm is used to train the database to achieve efficient, automatic, and fast classification of traffic signs. A dataset of traffic sign image recognition is used to simulate the experiment. The results show that the proposed Method improves the correct rate of intelligent recognition of traffic sign images, the recognition speed is greatly improved. Recognition accuracies of up to 96% were achieved in experiments on benchmark datasets, with inference times significantly lower than those of previous methods. The intelligent recognition results of traffic sign images are considerably better than other current methods, which have higher practical application value.

Zhenlin Luo, Kebin Lu

Abstract This study investigates the impact of a hybrid intelligent algorithm, integrating cognitive computing and deep learning, on the dynamic adjustment of employee performance evaluation in multi-scale organizational networks, through simulation experiments. Additionally, it proposes a performance evaluation mechanism based on algorithm optimization. The experiment initially compared the hybrid intelligent algorithm with the traditional KPI method. The results revealed that the mean square error (MSE) of the hybrid algorithm was significantly lower than that of the KPI (Key Performance Indicators) method across all datasets, with a 43.5% improvement in accuracy. It demonstrated superior accuracy in processing multi-dimensional employee data. Additional experiments involving noise interference indicate that the hybrid algorithm exhibits strong adaptability across varying data volumes. As the data size increases, the performance of the hybrid algorithm remains stable and continues to improve, outperforming traditional KPI and classical algorithms. Simultaneously, hybrid intelligent algorithms outperform support vector machines (SVM) in terms of response speed, with a 61.9% reduction in response time compared to SVM, highlighting their advantages in processing large-scale datasets. In terms of fairness, the hybrid intelligent algorithm outperforms the random forest algorithm (Gini coefficient of 0.22), with a lower Gini coefficient of 0.18, effectively reducing assessment bias and ensuring a fairer performance evaluation. Additionally, hybrid intelligent algorithms exhibit outstanding performance in improving employee satisfaction, with an 18.4% increase compared to traditional decision tree algorithms, suggesting that they provide more personalized feedback and enhance employees' identification with the performance appraisal system.

Solange MWAMBA NDUBA

Résumé : Cette étude avait pour but d’évaluer le degré de satisfaction des abonnés de téléphonie cellulaire face à la qualité de la connexion internet fournie par les opérateurs de la téléphonie mobile de la RDC. Pour ce faire, une échelle d’évaluation de la qualité de la connexion internet a été soumise à 108 étudiants finalistes du premier cycle de la Faculté de psychologie et des Sciences de l’Education de l’Université de Kinshasa. Les résultats obtenus ont indiqué que les sujets enquêtés sont globalement satisfaits de la qualité de la connexion internet, et cela dans toutes ses dimensions. Mots-clés : Satisfaction, Réseaux de téléphonie cellulaire, Qualité de la connexion, Coût de la connexion Internet, abonnés.

Sweta Agrawal, Amin Farajian, Patrick Fernandes et al.

Aditya Sivakumar, Ben Gelman, Robert Simmons

Abstract With the increasing availability of commercial Artificial Intelligence, General Language Models (GLMs) have been widely explored in various domains, including law. However, to ensure accurate and standardized legal results, it is crucial to establish a consistent framework for prompting GLMs. This paper presents one of the first instances of such nomenclature, providing a robust framework of “variables” and “clauses” that enhances legal-focused results. The proposed framework was applied in diverse legal scenarios, demonstrating its potential from both client and attorney perspectives. By introducing standardized variables and clauses, legal professionals can effectively communicate with GLMs. This not only improves the accuracy of the generated outcomes but also facilitates collaboration between AI systems and legal experts. With a common framework in place, legal practitioners can leverage AI technology confidently, knowing that the results produced align with established legal principles. Furthermore, the framework serves as a foundation for future research in the field of legal prompting with GLMs, and several avenues for future research are recommended in this paper. This standardization of nomenclature is expected to contribute to the wider adoption and benefit of GLMs in the legal field, leading to more accurate and reliable outcomes.

S. Varshini, Gottimukkala Sarayu Varma, S. M.

The scientific and technical field of computational linguistics seeks to comprehend spoken and written language from a computational standpoint. The way of describing rules and semantics in linguistics paved the beginning of natural language processing research for various languages spoken in the world. Over 700 languages are spoken in India alone, out of an estimated 7,000 spoken worldwide. Telugu is one of the most predominantly spoken languages in the states of Andhra Pradesh and Telangana. This proposed work presents a syntactical parsing technique on some basic sentences in Telugu. The Cocke-Younger-Kasami algorithm has been implemented to parse these basic sentences and also infer their grammatical structure. At present, there are very few language processing tools for Indian languages. Hence, an effort has been made to efficiently parse a few simple sentences in Telugu. The syntactical parser that has been developed acts as a recognizer and parser which can not only recognize and parse the grammatically correct sentences but can also recognize the grammatically incorrect sentences. This recognizer cum parser is then evaluated using the performance metrics like accuracy, precision and recall.

Md. Habibullah, Md. Shymon Islam, Fatima Tuz Jahura et al.

Massive digital texts are now accessible, thanks to technological advancement. Any amount of disorganized writing is useless. A high-quality representative corpus of any particular language is essential for research in computational linguistics and natural language processing (NLP). Bangla NLP research is still in its infancy because of the dearth of high-quality public corpus. This paper proposed a newly produced corpus consists of 1,30,307 documents covering 10 categories collected from 11 websites, having 2,94,80,828 tokens and 17,59,085 unique tokens. Seven supervised machine learning methods are explored in this work. Furthermore, Local Interpretable Model-agnostic Explanations (LIME) and SHapley Additive explanations (SHAP) are also examined to explain about different model performance. The obtained results show that the Random Forest (RF), Decision Tree (DT) and Support Vector Machine (SVM) outperform other models. RF classifier achieves the highest accuracy 99.91% which is better than the existing state-of-the-art methods.

Hailong Ning, Tao Lei, Mengyuan An et al.

Abstract Aerial scene recognition (ASR) has attracted great attention due to its increasingly essential applications. Most of the ASR methods adopt the multi‐scale architecture because both global and local features play great roles in ASR. However, the existing multi‐scale methods neglect the effective interactions among different scales and various spatial locations when fusing global and local features, leading to a limited ability to deal with challenges of large‐scale variation and complex background in aerial scene images. In addition, existing methods may suffer from poor generalisations due to millions of to‐be‐learnt parameters and inconsistent predictions between global and local features. To tackle these problems, this study proposes a scale‐wise interaction fusion and knowledge distillation (SIF‐KD) network for learning robust and discriminative features with scale‐invariance and background‐independent information. The main highlights of this study include two aspects. On the one hand, a global‐local features collaborative learning scheme is devised for extracting scale‐invariance features so as to tackle the large‐scale variation problem in aerial scene images. Specifically, a plug‐and‐play multi‐scale context attention fusion module is proposed for collaboratively fusing the context information between global and local features. On the other hand, a scale‐wise knowledge distillation scheme is proposed to produce more consistent predictions by distilling the predictive distribution between different scales during training. Comprehensive experimental results show the proposed SIF‐KD network achieves the best overall accuracy with 99.68%, 98.74% and 95.47% on the UCM, AID and NWPU‐RESISC45 datasets, respectively, compared with state of the arts.

Ignace KOFFI

Résumé : Face à la pauvreté grandissante des populations en cette période de crise militaro-politique qui a vu la partition du pays en deux zones administratives, depuis la nuit du 18 septembre 2002, des structures de financement, prônant la fortune par des excédents très aisés firent leur apparition en terre ivoirienne. L’Ivoirien qui ne sait pas être pauvre mais qui ne sait pas comment devenir riche va se lancer dans ce commerce de e-business. Les maisons de placement d’argent venaient de faire leur entrée en terre ivoirienne, promettant la fortune à tous par le jeu subtil de tontines. Mais quelle ne fut pas la désillusion des souscripteurs lorsque de façon spontanée et orchestrée, ces banques volantes disparurent spontanément et miraculeusement, faisant des clients, des victimes inconsolables. L’objectif de cette étude est de présenter les effets désastreux d’une crise qui a mis en branle l’économie et l’unité du pays. Il dévoile également les contrecoups de cette crise sur le vécu des populations. Cette étude relative à la manipulation monétaire présente d’abord les fondements de la banque en Côte d’Ivoire avant d’aborder le problème des banques volantes. Vouloir en faire le point sur le sujet ne serait pas approprié en ce moment car malgré les déconvenues, de nouvelles structures de placement d’argent ont vu le jour et continuent de faire des victimes. Cet article expose les résultats d’un système d’arnaque savamment agencé et conçu pour asservir des populations déjà fragilisées par la guerre. Mots clés : pauvreté, financement, e-business, fortune, victime.

Pengpeng Liu, Guixuan Zhang, Shuwu Zhang et al.

Abstract The aim is to reconstruct a complete and detailed clothed human from a single‐view input. Implicit function is suitable for this task because it represents fine shape details and varied topology. Current methods, however, often suffer from artefacts such as broken or disembodied body parts, missing details, or depth ambiguity due to the ambiguity and complexity of human articulation. The main issue observed by the authors is structure‐agnostic. To address these problems, the authors fully utilise the skinned multi‐person linear (SMPL) model and propose a method using the Skeleton‐aware Implicit Function (SIF). To alleviate the broken or disembodied body parts, the proposed skeleton‐aware structure prior makes the skeleton awareness into an implicit function, which consists of a bone‐guided sampling strategy and a skeleton‐relative encoding strategy. To deal with the missing details and depth ambiguity problems, the authors’ body‐guided pixel‐aligned feature exploits the SMPL to enhance 2D normal and depth semantic features, and the proposed feature aggregation uses the extra geometry‐aware prior to enabling a more plausible merging with less noisy geometry. Additionally, SIF is also adapted to the RGB‐D input, and experimental results show that SIF outperforms the state‐of‐the‐arts methods on challenging datasets from Twindom and Thuman3.0.

Сергій Різник

До 100-річчя з дня народження Іллі Корнійовича Кучеренка   Інформація про автора: Різник Сергій Михайлович – кандидат філологічних наук, доцент, завідувач кафедри української мови та прикладної лінгвістики Навчально-наукового інституту філології Київського національного університету імені Тараса Шевченка (Україна). Електронна адреса: serhiiriznyk2016@gmail.com

Anne Lauscher, B. Ko, Bailey Kuehl et al.

Citation context analysis (CCA) is an important task in natural language processing that studies how and why scholars discuss each others’ work. Despite decades of study, computational methods for CCA have largely relied on overly-simplistic assumptions of how authors cite, which ignore several important phenomena. For instance, scholarly papers often contain rich discussions of cited work that span multiple sentences and express multiple intents concurrently. Yet, recent work in CCA is often approached as a single-sentence, single-label classification task, and thus many datasets used to develop modern computational approaches fail to capture this interesting discourse. To address this research gap, we highlight three understudied phenomena for CCA and release MULTICITE, a new dataset of 12.6K citation contexts from 1.2K computational linguistics papers that fully models these phenomena. Not only is it the largest collection of expert-annotated citation contexts to-date, MULTICITE contains multi-sentence, multi-label citation contexts annotated through-out entire full paper texts. We demonstrate how MULTICITE can enable the development of new computational methods on three important CCA tasks. We release our code and dataset at https://github.com/allenai/multicite.

Nora Hollenstein, Emmanuele Chersoni, Cassandra L. Jacobs et al.

Eye-tracking data from reading represent an important resource for both linguistics and natural language processing. The ability to accurately model gaze features is crucial to advance our understanding of language processing. This paper describes the Shared Task on Eye-Tracking Data Prediction, jointly organized with the eleventh edition of the Work- shop on Cognitive Modeling and Computational Linguistics (CMCL 2021). The goal of the task is to predict 5 different token- level eye-tracking metrics of the Zurich Cognitive Language Processing Corpus (ZuCo). Eye-tracking data were recorded during natural reading of English sentences. In total, we received submissions from 13 registered teams, whose systems include boosting algorithms with handcrafted features, neural models leveraging transformer language models, or hybrid approaches. The winning system used a range of linguistic and psychometric features in a gradient boosting framework.

M. Misuraca, Alessia Forciniti, Germana Scepi et al.

Sentiment Analysis (SA) refers to a family of techniques at the crossroads of statistics, natural language processing, and computational linguistics. The primary goal is to detect the semantic orientation of individual opinions and comments expressed in written texts. There are several practical applications of SA in several domains. In an educational context, the use of this approach allows processing students' feedback, aiming at monitoring the teaching effectiveness of instructors and enhancing the learning experience. This paper wants to review the different R packages that can be used to carry on SA, comparing the implemented methods, discussing their characteristics, and showing how they perform by considering a simple example.

Yonatan Belinkov, Sebastian Gehrmann, Ellie Pavlick

While deep learning has transformed the natural language processing (NLP) field and impacted the larger computational linguistics community, the rise of neural networks is stained by their opaque nature: It is challenging to interpret the inner workings of neural network models, and explicate their behavior. Therefore, in the last few years, an increasingly large body of work has been devoted to the analysis and interpretation of neural network models in NLP. This body of work is so far lacking a common framework and methodology. Moreover, approaching the analysis of modern neural networks can be difficult for newcomers to the field. This tutorial aims to fill this gap and introduce the nascent field of interpretability and analysis of neural networks in NLP. The tutorial will cover the main lines of analysis work, such as structural analyses using probing classifiers, behavioral studies and test suites, and interactive visualizations. We will highlight not only the most commonly applied analysis methods, but also the specific limitations and shortcomings of current approaches, in order to inform participants where to focus future efforts.

Sai Ashish Somayajula, Linfeng Song, Pengtao Xie

AbstractText augmentation is an effective technique in alleviating overfitting in NLP tasks. In existing methods, text augmentation and downstream tasks are mostly performed separately. As a result, the augmented texts may not be optimal to train the downstream model. To address this problem, we propose a three-level optimization framework to perform text augmentation and the downstream task end-to- end. The augmentation model is trained in a way tailored to the downstream task. Our framework consists of three learning stages. A text summarization model is trained to perform data augmentation at the first stage. Each summarization example is associated with a weight to account for its domain difference with the text classification data. At the second stage, we use the model trained at the first stage to perform text augmentation and train a text classification model on the augmented texts. At the third stage, we evaluate the text classification model trained at the second stage and update weights of summarization examples by minimizing the validation loss. These three stages are performed end-to-end. We evaluate our method on several text classification datasets where the results demonstrate the effectiveness of our method. Code is available at https://github.com/Sai-Ashish/End-to-End-Text-Augmentation.

Irati De Nicolás , Ariane Sande Piñeiro

Due to the global COVID-19 pandemic, educational institutions all over the world were forced to make the executive decision of quickly and effectively transferring face-to-face classes into an online format. Issues such as technical difficulties, varying degrees of digital literacy within a given team of teachers or the question of how to preserve the essence of face-to-face classes in an online environment are among the many challenges that have affected and hindered the practice of instructors all over the world. With a focus on the teaching of Spanish as a Foreign Language (Español como Lengua Extranjera, ELE) and video-conferencing platforms such as the software known as Zoom, in this paper we discuss the importance of continuing the practice of teaching speaking in an online environment. Subsequently, we present a number of strategies that we have put into practice in our online ELE classrooms in order to try and maintain the same or very similar level of interaction between our students when compared to face-to-face classrooms.

Valerio Basile

L. Abualigah, Hamza Essam Alfar, Mohammad Shehab et al.