iosr-JCE   Volume-27 ~ Issue-2 ~ Mar. - Apr. 2025

Abstract: Matching resumes to job descriptions (JDs) is a critical process in personalized job search platforms, ensuring that candidates are effectively aligned with suitable opportunities. This study evaluates three key models—BERT, Gemini, and LLaMA 3.1—focusing on their capabilities, limitations, and comparative performance in addressing challenges such as semantic matching, domain-specific context handling, and variability in text representation. Comprehensive testing reveals insights into each model's strengths and weaknesses, offering a clear roadmap for selecting the optimal solution for AI-driven recruitment platforms. Our findings underscore LLaMA 3.1's superiority, highlighting its advanced contextual understanding, robust domain-specific matching capabilities, and ability to handle ambiguous resumes, establishing it as the most effective model for this application.

Keywords: Resume, Job Description, BERT, Gemini, LLaMA.

[1] Devaraju, Sudheer. (2022). Natural Language Processing (NLP) In AI-Driven Recruitment Systems. International Journal Of Scientific Research In Computer Science Engineering And Information Technology. 8. 555-566. 10.32628/CSEIT2285241.
[2] Maitra, Mayukh & Sinha, Surabhi & Kierszenowicz, Tomas. (2024). An Improved BERT Model For Precise Job Title Classification Using Job Descriptions. 1-6. 10.1109/IS61756.2024.10705204.
[3] Prathima, V.. (2024). Resume Application Tracking System With Google Gemini Pro. International Journal For Research In Applied Science And Engineering Technology. 12. 66-70. 10.22214/Ijraset.2024.59645.
[4] Bhatt, A., Vaghela, N., & Dudhia, K. (2024). Generating Knowledge Graphs From Large Language Models: A Comparative Study Of GPT-4, Llama 2, And BERT. Https://Doi.Org/10.48550/Arxiv.2412.07412.
[5] Musleh, D. A. (2024). Rule-Based Information Extraction From Multi-Format Resumes For Automated Classification. Mathematical Modelling Of Engineering Problems, 11(4), 1044–1052. Https://Doi.Org/10.18280/Mmep.110422.

Abstract: Bee sounds are an essential data source for monitoring beehives as they contain a wealth of helpful information related to the health and behavior of the bee colony. Numerous studies have reported that specific buzzing sounds depend on situations such as swarming, the absence of a queen, or detecting toxins in the air. In practice, many experienced beekeepers listen to their hives to assess the status of their honeybee colonies. A growing body of research has been focused on developing systems and methods to analyze and classify bee sound data to support beekeepers in improving the effectiveness of hive monitoring. Among these studies, machine learning (ML)- based methods are considered the most powerful techniques........

Keywords: Bee sounds; Feature extraction; MFCC; STFT;

[1] Davis S. Và Mermelstein, P. Comparison Of Parametric Representations For Monosyllabic Word Recognition In Continuously Spoken Sentences, IEEE Transactions On Acoustics, Speech, And Signal Processing, 28(4): 357-366, 1980.
[2] Nguyễn Trung Thành. MFCC Cho Xử Lý Tiếng Nói. Https://Viblo.Asia/P/Feature-Extraction-Mfcc-Cho-Xu-Ly-Tieng-Noi-4dbzn2xmzym, 2020.
[3] S. Cecchi Et Al., "Multi-Sensor Platform For Real Time Measurements Of Honey Bee Hive Parameters," IOP Conf. Ser.: Earth Environ. Sci., Vol. 275, P. 012016, May 2019, Doi: 10.1088/1755-1315/275/1/012016..
[4] V. Kulyukin And S. Mukherjee, "On Video Analysis Of Omnidirectional Bee Traffic: Counting Bee Motions With Motion Detection And Image Classification," Applied Sciences, Vol. 9, No. 18, P. 3743, Sep. 2019, Doi: 10.3390/App9183743.
[5] Breiman, L., 2001. Random Forests. Machine Learning 45, 5–32.

Abstract: Motion blur greatly impairs visual quality and diminishes the performance of subsequent computer vision tasks. This paper introduces a real-time selective motion deblurring system that adaptively detects and recovers solely motion-blurred areas in an image while keeping previously sharp areas intact. The system utilizes a hybrid motion blur detection module that combines Laplacian variance, FFT-based frequency analysis, and edge density estimation to produce a blur probability map. This image is divided with SLIC superpixels and post-processed using guided filtering........

Keywords: Motion Deblurring, Selective Deblurring, Real-Time Image Restoration, Edge-Preserving Attention, Adaptive Spatial Filtering, Swin Transformer, Superpixel Segmentation, Guided Filtering, MIMO-UNet, Hybrid Loss Function, FFT Blur Detection, CNN-Transformer Hybrid Models, Deep Learning in Vision.

[1]. R. Fergus, B. Singh, A. Hertzmann, S. Roweis, And W. T. Freeman, "Removing Camera Shake From A Single Photograph," Ieee Trans. Pattern Anal. Mach. Intell., Vol. 27, No. 8, Pp. 1154–1166, 2005.
[2]. S. Nah, T. H. Kim, And K. M. Lee, "Deep Multi-Scale Convolutional Neural Network For Dynamic Scene Deblurring," Ieee Trans. Pattern Anal. Mach. Intell., Vol. 41, No. 8, Pp. 1875–1888, 2019.
[3]. X. Wang, J. Chen, K. Yu, Et Al., "Edvr: Video Restoration With Enhanced Deformable Convolutional Networks," Ieee Trans. Pattern Anal. Mach. Intell., Vol. 43, No. 7, Pp. 2345–2360, 2021.
[4]. L. Sun, J. Chen, X. Wei, Et Al., "Eavt: Efficient Axial-Shifted Transformers For Video Deblurring," Ieee Trans. Pattern Anal. Mach. Intell., Vol. 43, No. 12, Pp. 4352–4364, 2021.
[5]. Y. Zhang, J. Xie, Et Al., "Dasnet: Degradation-Aware Selective Deblurring Network," In Proc. Ieee Conf. Comput. Vis. Pattern Recognit., 2021, Pp. 1523–1531.