خوشي سليم, دانية علي حسين, عمرسرمد طه

Year: 2025

Academic Degree: Bachelor

Supervisor Type: Supervisor

Supervisor State: Ungraduated

Development of a Simple Knowledge Graph: Computer Engineering Department as a Case Study

حسين طارق, عمر سرمد, محمد ازاد

Year: 2025

Academic Degree: Bachelor

Supervisor Type: Supervisor

Supervisor State: Ungraduated

Utilizing LLM to Identify Entities in Natural Language Text

B.Eng Technical Computer Engineering

Sep 5, 1999 - Aug 4, 2003

MSc. Networks Management

Oct 1, 2006 - Dec 12, 2007

Image-Based Text Translation a Fine-Tuning Using DeepSeek-Coder and Transformer Models for Multilingual Optical Character Recognition Processing.

Jul 1, 2025

Journal Mathematical Modelling of Engineering Problems

publisher IIETA

DOI 10.18280/mmep.120735

Issue 7

Volume 12

The combination of Optical Character Recognition (OCR) and deep learning-based translation has significantly enhanced multilingual text processing from images, especially. DeepSeek-Coder is a 1.3-billion-parameter autoregressive transformer model fine-tuned as a baseline for text translation from images in this study. Its performance is implemented and compared with other state-of-the-art transformer models Phi-1, MarianMT, MBart, M2M-100 and T5-Tiny. This work has prepared a dataset consisting of 1,000 images and their correlated English and French text which were extracted using EasyOCR. The performance is measured by using standard translation scores such as BLEU, METEOR, ROUGE-L, TER and perplexity. The DeepSeek-Coder achieves the best performance among other approaches with a BLEU score of 0.7733 and a very low perplexity of 1.28, beating all other models by far. These results demonstrate the outstanding translation accuracy and smoothness and highlight efficiency of transformer in OCR based translation. This study provides recommendations for the proper selection of a transformer model for processing of text have been emphasized through these findings. Additionally, the contributions of this work have been added to state-of-the-art in developing multilingual AI for real-life translation work with an OCR-based.

Read Publication

Hallucinations in GPT-2 Trained Model

Jan 25, 2025

Journal Ingénierie des Systèmes d’Information

publisher IIETA

DOI https://doi.org/10.18280/isi.300104

Issue 1

Volume 30

This paper analysis the phenomenon of "hallucinations" in text generated by GPT-2 when it produces irrelevant or illogical content. This work has quantified the extent of those hallucinations and look into ways of their mitigation. By using two main techniques: cosine similarity and frequency analysis. These techniques calculate coherency and relevance in the text produced by OpenAI GPT-2 at different training levels. Where a study case was implemented to train the model and ask the questions and retrain the model using these replays. The main findings indicate that this model hallucinates much less at the beginning of learning, with the situation significantly improving as training progresses. Extreme learning does not eliminate all such inadequacies, and more over-training led to more hallucinations. The hallucinated items span from smaller deviations to major content-wise deviations. An inspection reveals some patterns and cues that are predictive of increased output unreliability of the model. This research suggests a stricter training program that involve varied data sets to reduce the rate of hallucinations. More importantly, improve the accuracy of the model by reaching superior levels through the embedding of contextual and factual anchoring systems as well as designing algorithms for higher trigger identification. Other recommendations of the paper include post-generation text evaluation and continuous research to enhance the complexity of the models.

Read Publication

21st International Conference on Semantic Systems

Sep 3, 2025 - Sep 5, 2025

Country Austria

Location Vienna

Visit Conference

24th International Conference on Knowledge Engineering and Knowledge Management (EKAW 2024)

Nov 24, 2024 - Nov 28, 2024

Country Netherlands

Location Amsterdam

Visit Conference