YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Research Articles by Furkan Gözükara

A curated collection of research articles and theses by Furkan Gözükara and collaborators, spanning 2012-2025.

At a Glance

  • 10 works across journal articles, an MSc thesis, and a PhD thesis
  • Core themes: product search, record linkage, focused web crawling, sentiment analysis, cyber forensics, and human-computer interaction
  • Includes both method papers and full-system theses that connect crawling, normalization, matching, ranking, and evaluation

Research Themes

  • E-commerce search, comparison shopping, and product intelligence
  • Product identity clustering, record linkage, and noisy-data normalization
  • Focused web crawling and large-scale data extraction
  • Sentiment analysis for Turkish and English text
  • Cyber forensics and evidentiary risk analysis
  • Air-writing recognition and human-computer interaction

Quick Index

Year Title PDF Type Venue / Source Focus
2025 Letter and Person Recognition in Freeform Air-Writing Using Machine Learning Algorithms PDF Journal article IEEE Access, Vol. 13 Air-writing, person recognition
2021 An Incremental Hierarchical Clustering Based System For Record Linkage In E-Commerce Domain PDF Journal article The Computer Journal (uploaded PDF is an advance-article version) Record linkage, product matching
2021 Challenges and Possible Severe Legal Consequences of Application Users Identification from CNG-Logs PDF Journal article Forensic Science International: Digital Investigation, Vol. 39 CGNAT / cyber forensics
2017 Efficient Feature Selection for Product Labeling over Unstructured Data PDF Journal article IJACSA, Vol. 8, No. 7 Feature selection, clustering
2017 Focused Web Crawler Development Challenges: ECCrawler PDF Journal article International Journal of Computer Science and Engineering, Vol. 6, Issue 1 Focused crawling, systems engineering
2016 An Experimental Investigation of Document Vector Computation Methods for Sentiment Analysis of Turkish and English Reviews PDF Journal article Çukurova University Journal of the Faculty of Engineering and Architecture, Vol. 31, No. 2 Sentiment analysis
2016 A Product Search Engine Supporting "Best Product" Queries PDF Journal article Çukurova University Journal of the Faculty of Engineering and Architecture, Vol. 31, Special Issue 2 Product ranking, query processing
2016 Product Search Engine Using Product Name Recognition and Sentiment Analysis PDF PhD thesis Çukurova University Full product-search-engine architecture
2015 New Metrics for Clustering of Identical Products over Imperfect Data PDF Journal article Turkish Journal of Electrical Engineering and Computer Sciences, Vol. 23, No. 4 Similarity metrics, evaluation
2012 Fiyat Karşılaştırmalı Ürün Arama Motoru Geliştirme PDF MSc thesis Mersin University Price-comparison search engine

Detailed Timeline

2025 - Letter and Person Recognition in Freeform Air-Writing Using Machine Learning Algorithms

Type: Journal article
Venue: IEEE Access, Vol. 13
Focus: Air-writing, letter recognition, person recognition, IMU-based interaction

This paper introduces a wearable-glove pipeline for freeform air-writing analysis that jointly models letter recognition and writer recognition. It uses IMU signals, Fourier and wavelet feature extraction, and multiple machine-learning baselines, while also contributing a public Turkish alphabet air-writing dataset. The study reports that SubSpace KNN performs best under the tested settings.

2021 - An Incremental Hierarchical Clustering Based System For Record Linkage In E-Commerce Domain

Type: Journal article
Venue: The Computer Journal (the uploaded PDF is an advance-article version dated 2021 rather than a later issue-formatted PDF)
Focus: Record linkage, incremental clustering, product-title matching

This work presents a dynamic / incremental Hierarchical Agglomerative Clustering (HAC) system for grouping identical products crawled from different e-commerce websites. The method uses bag-of-words title representations, domain-specific matching / filtering, and ELKI-based evaluation, and reports 96.25% F-measure on the experimental setup. The paper also emphasizes dataset release and evaluation reproducibility.

2021 - Challenges and Possible Severe Legal Consequences of Application Users Identification from CNG-Logs

Type: Journal article
Venue: Forensic Science International: Digital Investigation, Vol. 39
Focus: CGNAT, reverse tracking, cyber forensics, evidentiary risk

This paper studies how carrier-grade NAT / CGNAT logs can be misused in reverse-tracking workflows and how such misuse can lead to false attribution in criminal investigations. Using the ByLock case context in Turkey and a comparison with EncroChat, it analyzes the technical and legal consequences of flawed identification pipelines.

2017 - Efficient Feature Selection for Product Labeling over Unstructured Data

Type: Journal article
Venue: International Journal of Advanced Computer Science and Applications (IJACSA), Vol. 8, No. 7
Focus: Feature selection, product labeling, clustering under unstructured data

This study proposes a feature-selection algorithm for labeling identical products collected from noisy, heterogeneous web sources. The paper frames product labeling as a clustering problem over unstructured feature vectors and shows that the proposed method improves clustering quality compared with baseline approaches.

2017 - Focused Web Crawler Development Challenges: ECCrawler

Type: Journal article
Venue: International Journal of Computer Science and Engineering, Vol. 6, Issue 1
Focus: Focused crawling, multithreading, .NET systems engineering

This paper documents the engineering of EcCrawler, a hand-crafted focused crawler for e-commerce websites built with C#, .NET 4.5, and MS-SQL Server 2014. It focuses on practical implementation topics such as threading, exception handling, HTTP compression, duplicate handling, and database communication, and reports over 400% crawling-speed improvement and over 100% UI-responsiveness improvement from the proposed optimizations.

2016 - An Experimental Investigation of Document Vector Computation Methods for Sentiment Analysis of Turkish and English Reviews

Type: Journal article
Venue: Çukurova University Journal of the Faculty of Engineering and Architecture, Vol. 31, No. 2
Focus: Sentiment analysis, vectorization, feature selection, Turkish and English reviews

This article compares document-vector construction choices for sentiment analysis, including TF / TF-IDF variants, tokenization, feature selection, preprocessing, and vector normalization under an SVM classifier. On the collected Turkish product-reviews dataset, it reports a best result of 91.33% accuracy.

2016 - A Product Search Engine Supporting "Best Product" Queries

Type: Journal article
Venue: Çukurova University Journal of the Faculty of Engineering and Architecture, Vol. 31, Special Issue 2
Focus: Product ranking, comparison shopping, query processing

This work presents a product-search-engine system that supports "find the best products for a given category" queries. The system integrates a focused crawler, record linkage, sentiment analysis, and a query engine, and reports 96.25% F-measure in record linkage together with 100% precision in the evaluated most-related-products search setting.

2016 - Product Search Engine Using Product Name Recognition and Sentiment Analysis

Type: PhD thesis
Institution: Çukurova University, Department of Computer Engineering
Focus: End-to-end product search engine architecture

This dissertation brings the main threads of the repository together into a full product search engine: focused crawling, product-name matching / record linkage, sentiment analysis, and a user-facing search system. The abstract reports 472% crawler performance boost, 91.08% sentiment-analysis accuracy, 96.25% F-measure for record linkage, and 100% precision for most-related-products search in the thesis setup.

2015 - New Metrics for Clustering of Identical Products over Imperfect Data

Type: Journal article
Venue: Turkish Journal of Electrical Engineering and Computer Sciences, Vol. 23, No. 4
Focus: Similarity metrics, performance metrics, imperfect web-crawled product data

This paper formalizes product identity-clustering for web-crawled commercial products described by noisy, incomplete, and structurally inconsistent data. It proposes new similarity metrics and new evaluation metrics for this setting and shows that legacy measures such as Euclidean and cosine similarity are weaker on the tested product-clustering problem.

2012 - Fiyat Karşılaştırmalı Ürün Arama Motoru Geliştirme

Type: MSc thesis
Institution: Mersin University, Department of Computer Engineering
Focus: Price-comparison search, normalization, feature extraction, clustering

This master's thesis lays the early foundation for a price-comparison product search engine. It covers focused collection of product data, noise removal / normalization, feature-vector extraction, and clustering of identical products across sources, and it also includes an English abstract under the title "Developing Product Price Comparison Search Engine."

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support