466 Citations8 h-index7 i10-indexSource: OpenAlex · Google Scholar →
No publications found.
2025
Web Kazıma: Disiplinler Arası Veri Toplama ve Açık Kaynak Çözümler
Büyük Dil Modelleri ve yapay zekâdaki gelişmeler, web ajanlarında kural tabanlı geleneksel yaklaşımlardan, kullanıcı niyetlerini anlayıp dinamik ortamlarda otonom kararlar alabilen akıllı ajanlara doğru güçlü bir paradigma değişimi yaratmıştır. Bu kitap bölümü, otonom web ajanlarının teorik temellerini, farklı yaklaşımları ve performans değerlendirmeleri bütüncül bir bakış açısıyla ele almaktadır. Çalışma kapsamında, otonom web ajanlarının web tarayıcısı nasıl etkileşime girdiği, HTML/DOM tabanlı çözümler ile web sayfasının görseli üzerinden bir kullanıcı gibi hareket etmeye kadar olan çözümler karşılaştırılmalı olarak ele alınmıştır. Ayrıca, web ajanlarının hiyerarşik planlamalar yapmasına olanak sağlayan ReAct gibi akıl yürütme algoritmalarının nasıl çalıştığı incelenir. Web ajanlarının performanslarını değerlendirme WebArena ve Mind2Web gibi literatürde kullanılan test ortamları ve başarı oranı, verimlilik ve maliyet gibi performans metriklerine yer verilir. Yapılan incelemeler, görsel tabanlı ve DOM tabanlı yöntemler yerine hibrit mimarilerin otonom web ajanlarında kullanılmasının daha başarılı sonuçlar verebildiğini göstermiştir. Günümüzde LLM’lerin büyük sorunu olan halüsinasyon problemi web ajanlarında da karşımıza çıkmaktadır. Bununla birlikte, verimliliğin düşük olması ve sonsuz döngü problemleri gibi ek sorunlar da ortaya çıkmaktadır. Kitap bölümünün son kısmında, verimsizliğin iyileştirilmesi, farklı dil modellerinin test edilmesi, zero-shot gibi literatürde bilinen yöntemlerin web ajanlarına taşınması, geri bildirimli web ajanları, web ajan güvenliği ve yaşam boyu öğrenme alanında gelecekte yapılabilecek araştırmalar önerilmiştir.Web Tarayıcılarının Yeni Kullanıcıları: Otonom Web Ajanlarının Yükselişi
Günümüzde, internet üzerindeki büyük ve sürekli güncellenen veri kümeleri sağlık, ekonomi, beşerî ve sosyoloji olmak üzere birçok alandaki bilimsel araştırma süreçlerine katkı sağlayabilecek hale gelmiştir. Özellikle, anket ve yüz yüze görüşme gibi geleneksel yöntemler yerine güncel ve geniş kapsamlı internet ortamındaki veri kümeleri araştırmacıların hipotezlerini oluşturup daha güvenilir sonuçlar almasını desteklemektedir. Bu kitap bölümü, bilgisayar bilimleri alanında uzmanlığı olmayan araştırmacılara, kendi veri kümelerini web sayfalarından toplamak için web kazıma süreçlerini nasıl gerçekleştiğini ve kullanabileceği açık kaynaklı ürünleri tanıtmaktadır. Bu çalışmada, web kazımanın temelini oluşturan internetin nasıl çalıştığı açıklanmış ve web sayfalarında kullanılan HTML yapıları hakkında bilgi verilmiştir. Ardından sorumlu veri toplamanın önemi etik kurallar ve hukuki boyut bağlanımda detaylandırılmıştır. Web kazıma uygulaması geliştirmek için kullanılabilecek tarayıcı eklentileri açıklandıktan sonra Python (Requests, Beautiful Soup, Pandas) ve dinamik siteler için Playwright/Selenium gibi açık kaynaklı çözümler örnek bir web sitesi üzerinden web kazıma işlemi yapılarak açıklanmıştır. Ayrıca, web kazıma aşamasında CAPTCHA, IP Engellenmesi, kural değişiklikleri gibi zorluklar ele alınmıştır. Web kazıma beceresi, birçok bilim alanı için özgün ve güncel çalışmalar yapmanın anahtarıdır. Web kazıma tekniklerini konu alan bu bölüm, farklı disiplinlerdeki araştırmacıların kendi özgün veri kümelerini oluşturarak, çalışmalarının deneysel dayanağını ve sonuçlarının doğruluğunu güçlendirmeyi hedeflemektedir.
2024
Scraping Relevant Images from Web Pages Without Download
Automatically scraping relevant images from web pages is an error-prone and time-consuming task, leading experts to prefer manually preparing extraction patterns for a website. Existing web scraping tools are built on these patterns. However, this manual approach is laborious and requires specialized knowledge. Automatic extraction approaches, while a potential solution, require large training datasets and numerous features, including width, height, pixels, and file size, that can be difficult and time-consuming to obtain. To address these challenges, we propose a semi-automatic approach that does not require an expert, utilizes small training datasets, and has a low error rate while saving time and storage. Our approach involves clustering web pages from a website and suggesting several pages for a non-expert to annotate relevant images. The approach then uses these annotations to construct a learning model based on textual data from the HTML elements. In the experiments, we used a dataset of 635,015 images from 200 news websites, each containing 100 pages, with 22,632 relevant images. When comparing several machine learning methods for both automatic approaches and our proposed approach, the AdaBoost method yields the best performance results. When using automatic extraction approaches, the best f-Measure that can be achieved is 0.805 with a learning model constructed from a large training dataset consisting of 120 websites (12,000 web pages). In contrast, our approach achieved an average f-Measure of 0.958 for 200 websites with only six web pages annotated per website. This means that a non-expert only needs to examine 1,200 web pages to determine the relevant images for 200 websites. Our approach also saves time and storage space by not requiring the download of images and can be easily integrated into currently available web scraping tools, because it is based on textual data.
2023
An efficient regular expression inference approach for relevant image extraction
Traditional approaches for extracting relevant images automatically from web pages are error-prone and time-consuming. To improve this task, operations such as preparing a larger dataset and finding new features are used in the web data extraction approaches. However, these operations are difficult and laborious. In this study, we propose a fully-automated approach based on alignment of regular expressions to automatically extract the relevant images from web pages. The automatically constructed regular expressions has been applied to a classification task for the first time. In this respect, a multi-stage inference approach is developed for generating regular expressions from the attribute values of relevant and irrelevant image elements in web pages. The proposed approach reduces the complexity of the alignment of two regular expressions by applying a constraint on a version of the Levenshtein distance algorithm. The classification accuracy of regular expression approaches is compared with the naive Bayes, logistic regression, J48, and multilayer perceptron classifiers on a balanced relevant image retrieval dataset consisting of 360 image element samples for 10 shopping websites. According to the cross-validation results, the regular expression inference-based classification achieved a 0.98 f-measure with only 5 frequent n-grams, and it outperformed other classifiers on the same set of features. The classification efficiency of the proposed approach is measured at 0.108 ms, which is very competitive with other classifiers.
2021
The Analysis of Big Financial Data Through Artificial Intelligence Methods
A new data world which never get deformed, can be reached from anywhere, continuously stream and multiply, emerged with the evolution of technology. The data, in particular, created by business firms, scientific research centers, and automation systems reached great amounts. It has become the main target of many data analysts to reach meaningful, unexplored, and valuable information or deductions among these piles of data. In this chapter, firstly the techniques of artificial intelligence and the skills of these techniques were discussed. Later, the mostly-used techniques in the finance sector, the advantages and weaknesses of these techniques, and the methods which can be used to process the data created by the finance sector, which creates big data and is one of the leading sources, was comparatively shown. The current version of the mostly-used artificial intelligence methods in the finance sector was scanned and the new skills and contributions it provides to the sector were examined. What Classification, clustering, association rules, and time series analysis methods, in particular, cover and what problems they can produce solutions to were examined and the readers were informed about these techniques. It was aimed to give information about forming credit score and customer segmentation, where classification and clustering methods are especially employed, with sample studies. It was aimed to present the principles the up-to-date methods are based on and their theoretical and practical applications in a meaningful way. In addition to these, information about practical and useful software that can be used for data analysis in the finance sector was given and the skills of this software were conveyed to the readers. Finally, how the techniques of processing big data can be used was examined through samples as the finance data are classified as big data. The difficulties met during the analysis of big data, a natural result created by this sector, and solutions to them were presented. Updated big data processing solutions like Hadoop, Spark, MapReduce, Distributed computing, and GPU (Graphics Processing Unit) computing, in particular, were comparatively explained. The main principles that big data processing techniques are based on were simplified in a way that the readers could understand and were supported by examples from the sector. Especially, Spark, Hadoop, and MapReduce methods, which are leading methods in processing big data, were examined. Finally, the contributions made to the sector by artificial intelligence and big data processing techniques were generally summarized and the results were presented.
2020
A Novel Web Scraping Approach Using the Additional Information Obtained from Web Pages
Web scraping is a process of extracting valuable and interesting text information from web pages. Most of the current studies targeting this task are mostly about automated web data extraction. In the extraction process, these studies first create a DOM tree and then access the necessary data through this tree. The construction process of this tree increases the time cost depending on the data structure of the DOM Tree. In the current web scraping literature, it is observed that time efficiency is ignored. This study proposes a novel approach, namely UzunExt, which extracts content quickly using the string methods and additional information without creating a DOM Tree. The string methods consist of the following consecutive steps: searching for a given pattern, then calculating the number of closing HTML elements for this pattern, and finally extracting content for the pattern. In the crawling process, our approach collects the additional information, including the starting position for enhancing the searching process, the number of inner tag for improving the extraction process, and tag repetition for terminating the extraction process. The string methods of this novel approach are about 60 times faster than extracting with the DOM-based method. Moreover, using these additional information improves extraction time by 2.35 times compared to using only the string methods. Furthermore, this approach can easily be adapted to other DOM-based studies/parsers in this task to enhance their time efficiencies.A regular expression generator based on CSS selectors for efficient extraction from HTML pages
Cascading style sheets (CSS) selectors are patterns used to select HTML elements. They are often preferred in web data extraction because they are easy to prepare and have short expressions. In order to be able to extract data from web pages by using these patterns, a document object model (DOM) tree is constructed by an HTML parser for a webpage. The construction process of this tree and the extraction process using this tree increase time and memory costs depending on the number of HTML elements and their hierarchies. For reducing these costs, regular expressions can be considered as a solution. However, preparing regular expression patterns is a laborious task. In this study, a heuristic approach, namely Regex Generator (REGEXN), that automatically generates these patterns through CSS selectors is introduced and th eperformance gains are analyzed on a web crawler. The analysis shows that regular expression patterns generated by this approach can significantly reduce the average extraction time results from 743.31 ms to 1.03 ms when compared with the extraction process from a DOM tree. Similarly, the average memory usage drops from 1054.01 B to 1.59 B. Moreover, REGEXN can be easily adapted to the existing frameworks and tools in this task.Automatically Discovering Relevant Images From Web Pages
Web pages contain irrelevant images along with relevant images. The classification of these images is an error-prone process due to the number of design variations of web pages. Using multiple web pages provides additional features that improve the performance of relevant image extraction. Traditional studies use the features extracted from a single web page. However, in this study, we enhance the performance of relevant image extraction by employing the features extracted from different web pages consisting of standard news, galleries, video pages, and link pages. The dataset obtained from these web pages contains 100 different web pages for each 200 online news websites from 58 different countries. For discovering relevant images, the most straightforward approach extracts the largest image on the web page. This approach achieves a 0.451 F-Measure score as a baseline. Then, we apply several machine learning methods using features in this dataset to find the most suitable machine learning method. The best f-Measure score is 0.822 using the AdaBoost classifier. Some of these features have been utilized in previous web data extraction studies. To the best of our knowledge, 15 new features are proposed for the first time in this study for discovering the relevant images. We compare the performance of the AdaBoost classifier on different feature sets. The proposed features improve the f-Measure by 35 percent. Besides, using only the cache feature, which is the most prominent feature, corresponds to 7 percent of this improvement.Use of Machine Learning Algorithms In Classification Of DNA Gene Sequences: An Empiric Research on The Prometer And Splice Junction Data Set
Artificial Intelligence technologies can provide effective solutions in understanding the increasingly complex multi-dimensional data and obtaining meaningful information. In particular, as the number of attributes of the data, the complexity and size of the problem increase, the discovery of meaningful relationships, comprehensibility and interpretation of the data become increasingly difficult. To overcome this difficulty, algorithms developed by machine learning methods, a sub-branch of artificial intelligence, can be useful. Classification of DNA gene sequences is one of the difficult problems to solve, especially in molecular biology. This study investigates the applicability of machine learning algorithms, a sub-branch of artificial intelligence methods, on two data sets that contain the DNA sequence analysis problem and the effects of new approaches on this problem. For this purpose, previously untested machine learning algorithms and tools were used and new findings were achieved. Especially, Auto-Weka tool, which was developed to determine algorithms that offer optimum performance together with their parameters, was found to be effective. The findings show that the performance rates previously obtained on this data set can be improved with new approaches. In particular, it has been observed that Random Forest and SMO (Sequential minimal optimization) algorithms significantly increase classification performance rates compared to previous studies with the determination of appropriate parameter settings. In this study, in addition to increasing the performance of classification algorithms, feature reduction method, which has not been tried in previous researches, has also been considered and it has been shown that the number of features can be reduced significantly to increase the performance rate. In particular, it has been observed that the CfsSubsetEval algorithm can detect the most important gene sequences that affect the classification by significantly reducing the input parameters. In particular, it has been observed that the CfsSubsetEval algorithm can detect the most important gene sequences that affect the classification by significantly reducing the input parameters in Prometer data with the GreedyStepwise search method. The same algorithm has managed to reduce the number of attributes in the Slipce Junction data set at the optimum level with the BestFirst search method. Thus, it is possible to use hardware resources such as computational requirements and storage in the processing of DNA gene sequences more efficiently and to get the result quickly.
2019
Predicting CO and NOx emissions from gas turbines: Novel data and a benchmark PEMS
Predictive emission monitoring systems (PEMS) are important tools for validation and backing up of costly continuous emission monitoring systems used in gas-turbine-based power plants. Their implementation relies on the availability of appropriate and ecologically valid data. In this paper, we introduce a novel PEMS dataset collected over five years from a gas turbine for the predictive modeling of the CO and NOx emissions. We analyze the data using a recent machine learning paradigm, and present useful insights about emission predictions. Furthermore, we present a benchmark experimental procedure for comparability of future works on the dataTurkish Datasets for Text Genre Identification
In this study, 7 big data sets have been created to be used in modeling studies which can be asolution to the genre identification problem from text classification problems for Turkishlanguage, and these data sets consist of all the past column writings of the authors of anewspaper until 24.04.2019. These data sets, which were created for the purpose ofdetermining the type of text of the authors, columns, 8 classes and 24,344 texts, 7 classes and22,442 texts; 6 classes and 20,160 texts; 5 classes and 19,325 texts; 4 classes and 16,328texts; includes 3 classes and 14,334 texts and 2 classes and 11,698 texts. These datasets wereapplied to Naive Bayes and Random Forest classification algorithms which are the classicalmachine learning algorithms. The most successful result was obtained from the RandomForest classifier with 95.45% accuracy with a data set consisting of 2 classes, 5,849 texts ineach class and a total of 11,698 texts. (PDF) Metin Türü Belirleme için Türkçe Veri Setleri Turkish Datasets for Text Genre Identification. Available from: https://www.researchgate.net/publication/338232364_Metin_Turu_Belirleme_icin_Turkce_Veri_Setleri_Turkish_Datasets_for_Text_Genre_Identification [accessed Nov 22 2026].
2018
Comparison of Python Libraries used for Web Data Extraction
(Selected Paper for Journal) There are several libraries for extracting useful data from web pages in Python. In this study, we compare three different well-known extraction libraries including BeautifulSoup, lxml and regex. The experimental results indicate that regex achieves the best results with an average of 0.071 ms. However, it is difficult to generate correct extraction rules for regex when the number of inner elements is not known. In experiments, only .5 of the extraction rules are suitable for this task. In this case, BeautifulSoup and lxml, which are the DOM-based libraries, are used for extraction process. In experiments, lxml library yields the best results with an average of 9.074 ms.Comparison of Python Libraries used for Web Data Extraction
There are several libraries for extracting useful data from web pages in Python. In this study, we compare three different well-known extraction libraries including BeautifulSoup, lxml and regex. The experimental results indicate that regex achieves the best results with an average of 0.071 ms. However, it is difficult to generate correct extraction rules for regex when the number of inner elements is not known. In experiments, only .5 of the extraction rules are suitable for this task. In this case, BeautifulSoup and lxml, which are the DOM-based libraries, are used for extraction process. In experiments, lxml library yields the best results with an average of 9.074 ms.Examining the Impact of Feature Selection on Classification of User Reviews in Web Pages
The user reviews in web pages can provide useful information about the content of the web page for text processing applications. Automatically extracting data from a web page is a crucial process for these applications. One of the used methods in this process is to construct a learning model with an appropriate classification method using features that are derived from data. However, some features can be either redundant or irrelevant for this model. In this study, an imbalanced dataset including 47 shallow text features obtained from web pages is utilized for extracting of the user reviews. Then, various well-known feature selection techniques are applied to reduce the number of these features. The effects of this reduction on the classification methods are also examined. The experimental results indicate that approximately half of the features are sufficient for the classification task. Additionally, the AdaBoost classifier gives the best results concerning precision of about 0.930 for the review layout prediction.Object-based Entity Relationship Diagram Drawing Library: EntRel.JS
(Selected paper for Journal) An entity relationship diagram (ERD) is a visual helper for database design. ERD gives information about the relations of entity sets and the logical structure of databases. In this paper, we introduce an open source JavaScript Library named EntRel.JS in order to design sophisticated ERDs by writing simple codes. This library, which we have developed to facilitate this task, is based on our obfc.js library. It generates a SVG output on the client side for modern browsers. The SVG output provides storage efficiency when compared to existing ERD drawings created with existing drawing applications. Moreover, we present our animation library to gain action for elements in your web page.Object-based Entity Relationship Diagram Drawing Library: EntRel.JS
An entity relationship diagram (ERD) is a visual helper for database design. ERD gives information about the relations of entity sets and the logical structure of databases. In this paper, we introduce an open source JavaScript Library named EntRel.JS in order to design sophisticated ERDs by writing simple codes. This library, which we have developed to facilitate this task, is based on our obfc.js library. It generates a SVG output on the client side for modern browsers. The SVG output provides storage efficiency when compared to existing ERD drawings created with existing drawing applications. Moreover, we present our animation library to gain action for elements in your web page.Performance Evaluation of Classification Methods in Layout Prediction of Web Pages
The Web is an invaluable source of data stored on web pages. These data are contained in HTML layout elements of a web page. It is a crucial issue to extract data automatically from a web page. In this study, a dataset, which is annotated with seven different layouts including main content, headline, summary, other necessary layouts, menu, link, and other unnecessary layouts, is used. Then, 49 different features are computed from these layouts. Finally, we compare the different classification methods for evaluating the performance of these methods in layout prediction. The experiments show that the Random Forest classifier achieves a high accuracy of 98.46%. Thanks to this classifier, the prediction of link layout has a higher performance (approximately 0.988 f-Measure) according to the performance of the prediction of other layouts. On the other hand, the prediction of the summary layout has the worst performance with about 0.882 f-Measure.Veritabanı Tasarımının Yazılım Performansına Etkisi: Normalizasyona karşı Denormalizasyon
Yazılım performansını etkileyen en önemli faktörlerden biri veritabanı tasarımında yapılabilecek iyileştirmelerdir. Veritabanı tasarımında sıklıkla ilişkisel veritabanı teorisi olan normalizasyon işlemi kullanılır. Fakat veri miktarı arttıkça normalizasyon işleminden kaynaklı performans sorunları ortaya çıkmaya başlar. Performans sorunlarını ortadan kaldırmak için teorisi oluşmamış denormalizasyon işlemi kullanılır. Bu çalışmada, bir anket uygulamasında performans arttırıcı bir veritabanı tasarımı tanıtılmış ve bu veritabanı tasarımının MySQL, PostgreSQL ve Oracle olmak üzere üç farklı ilişkisel veritabanı yönetim sistemindeki performans artışı incelenmiştir. Ayrıca, günümüzün popüler veritabanı sistemlerinden NoSQL’e ne zaman geçilmesi gerektiği CAP teoremi üzerinden anlatılıp, normalizasyon ve denormalizasyon işlemlerinin bu teoremdeki yeri belirtilmiş olacaktır.
2017
A novel algorithm for extracting the user reviews from web pages
Extracting the user reviews in websites such as forums, blogs, newspapers, commerce, trips, etc. is crucial for text processing applications (e.g. sentiment analysis, trend detection/monitoring and recommendation systems) which are needed to deal with structured data. Traditional algorithms have three processes consisting of Document Object Model (DOM) tree creation, extraction of features obtained from this tree and machine learning. However, these algorithms increase time complexity of extraction process. This study proposes a novel algorithm that involves two complementary stages. The first stage determines which HTML tags correspond to review layout for a web domain by using the DOM tree as well as its features and decision tree learning. The second stage extracts review layout for web pages in a web domain using the found tags obtained from the first stage. This stage is more time-efficient, being approximately 21 times faster compared to the first stage. Moreover, it achieves a relatively high accuracy of 96.67% in our experiments of review block extraction.Comparison of string matching algorithms in web documents
String matching algorithms try to find position/s where one or more patterns (also called strings) are occurred in text. In this study, we compare 31 different pattern matching algorithms in web documents. In web documents, searching is crucial process for content extraction process. Therefore, lengths of html tags are examined for determining which algorithm or algorithms are suitable for matching process. Our experiments show that Skip Search algorithm is the best pattern matching algorithm with 0.170 ms for web documents. Moreover, it has 0.002 ms in preprocessing time and 0.168 ms in searching time.Evaluation of Hap, AngleSharp and HtmlDocument in web content extraction
Object-based flowchart drawing library
While flow charts are one of the best ways to describe a computer program, drawing process is a laborious task for developers. This study describes an open source Javascript library named obfc.js, which we have developed to facilitate this task. This library generates SVG output on the client side for modern browsers and allows easy creation of diagrams and links. Moreover, it allows you to link click events to objects and links. This library will allow you to design sophisticated flow charts using a very small amount of text data instead of both large image and SVG Data.
2016
Hiyerarşik kümeleme modeli kullanan web tabanlı bir ödev değerlendirme sistemi
Assignmentsare one of the most important parts of education process of students. In theclassical assignment evaluation process, an assignment can be evaluated whetherit is correct or not. However, for the assignments to give better contributionto education, plagiarisms committed by students should be considered. Detectionof plagiarism and its extent are extremely difficult assignment evaluationprocedures. In this study, in order to facilitate this procedure, a web-basedapplication, which can combine document similarity measures with hierarchicalclustering model, is introduced. This application gives the opportunity toevaluate which students submit similar assignments and the assignments’similarity degree. Cosine, Dice and Jaccard similarity measures have beeninvestigated in terms of document similarity calculation of this application.On the other hand, three different algorithms including Single Linkage, CompleteLinkage and Average Group are examined in hierarchical clustering side. Testdata which covers two education period of previous years and contains 54different assignments of 18 different courses of 6 lecturers, are created. Byusing document similarity methods and hierarchical clustering algorithms, 9different cophenetic correlation coefficients are obtained for each assignmentand cophenetic correlation coefficients are calculated to test how wellhierarchical clustering algorithms are . When the results were analyzed, it wasdiscovered that Jaccard measure in document similarity and Average Groupalgorithm in hierarchical clustering is the best matching assignment evaluationpair.
2015
- Elektrik ve güneş enerjisi ile çalışan taşıtlar hareket enerjilerini elektrik motorları aracılığı ile almaktadır.Bu motorlara enerji ise bataryalar tarafından sağlanmaktadır.Elektrikli araçların batarya durumu, sıcaklığı ve anlık hızı gibi veriler kalan yol miktarı gibi bilgilerin hesaplanması için oldukça önemlidir.Bu çalışmada bu verilerin ölçülmesi ve uzaktan takip edilebilmesi için fiziksel programlama platformları kullanılarak bir telemetri sistemi geliştirilmiştir.Veriler çeşitli sensörler aracılığı ile ölçülmüş, fiziksel programlama platformu aracılığı ile yorumlanmış ve GSM şebekesi aracılığı ile internet üzerinden bir sunucuya gönderilip, geliştirilen bir web sitesi üzerinde yayınlanmıştır.
2014
An effective and efficient web content extractor for optimizing the crawling process
Classical Web crawlers make use of only hyperlink information in the crawling process. However, focused crawlers are intended to download only Web pages that are relevant to a given topic by utilizing word information before downloading the Web page. But, Web pages contain additional information that can be useful for the crawling process. We have developed a crawler, iCrawler (intelligent crawler), the backbone of which is a Web content extractor that automatically pulls content out of seven different blocks: menus, links, main texts, headlines, summaries, additional necessaries, and unnecessary texts from Web pages. The extraction process consists of two steps, which invoke each other to obtain information from the blocks. The first step learns which HTML tags refer to which blocks using the decision tree learning algorithm. Being guided by numerous sources of information, the crawler becomes considerably effective. It achieved a relatively high accuracy of 96.37% in our experiments of block extraction. In the second step, the crawler extracts content from the blocks using string matching functions. These functions along with the mapping between tags and blocks learned in the first step provide iCrawler with considerable time and storage efficiency. More specifically, iCrawler performs 14 times faster in the second step than in the first step. Furthermore, iCrawler significantly decreases storage costs by 57.10% when compared with the texts obtained through classical HTML stripping. Copyright © 2013 John Wiley & Sons, Ltd.
2013
A hybrid approach for extracting informative content from web pages
Author detection by using different term weighting schemes
In this study, the impact of term weighting on author detection as a type of text classification is investigated. The feature vector being used to represent texts, consists of stem words as features and their weight values, which are obtained by applying 14 different term weighting schemes. The performances of these feature vectors for 3 different datasets in the author detection are tested with some classification methods such as Naïve Bayes Multinominal (NBM), and Support Vector Machine (SVM), Decision Tree (C4.5), and Random Forrest (RF), and are compared with each other. As a result of that, the most successful classifier, which can predict the author of an article, is found as SVM classifier with 98.75% mean accuracy; the most successful term weighting scheme is found as ACTF.IDF.(ICF+1) with 91.54% general mean accuracy.
2012
A fuzzy ranking approach for improving search results in Turkish as an agglutinative language
Text classification of web based news articles by using Turkish grammatical features
The dimensions of the feature vectors being used at the classification methods in the literature affect directly the time performance. In this study, how to reduce the dimension of the feature vector by using Turkish's grammar rules without compromising success rates is explained. The feature vector is weighted on the basis of the word frequency as the word stems have been selected as features. During this selection the effects of selection of the word stems with different length and type to the classification are investigated and when the word stems with noun type and the maximum length are selected as features, the success rate has been found to be at the highest level. When this selection is applied with the other methods which reduce the dimension, the dimension of the feature vector is decreased to 97.46%. Using the reduced feature vector the better succes rates generally have been obtained from Naive Bayes, SVM, C 4.5 and RF classification methods and the best performance achieved is 92.73% which has been obtained using the Naive Bayes method.Web content extraction by using decision tree learning
Via information extraction techniques, web pages are able to generate datasets for various studies such as natural language processing, and data mining. However, nowadays the uninformative sections like advertisement, menus, and links are in increase. The cleaning of web pages from uninformative sections, and extraction of informative content has become an important issue. In this study, we present an decision tree learning approach over DOM based features which aims to clean the uninformative sections and extract informative content in three classes: title, main content, and additional information. Through this approach, differently from previous studies, the learning model for the extraction of the main content constructed on DIV and TD tags. The proposed method achieved 95.58% accuracy in cleaning uninformative sections and extraction of the informative content. Especially for the extraction of the main block, 0.96 f-measure is obtained.
2011
2010
2008
2007
2006
2004
A New Approach to Developing A Data Exchange Language as An Alternative to XML
E-Belediyecilik ve Bir Kent Bilgi Sistemi Uygulaması
As a result of the studies we have carried out in the field of E-Municipality, we have decided to focus on City Information Systems, which is a sub-field of E-Municipality, and solidified our study by developing an application. The application allows the user to have access to the information concerning a parcel by entering the Plot, Block and Parcel Numbers or using the map. We have chosen PHP as the programming language and MYSQL as the database for our application.
2003
E-Belediyecilik ve Bir Kent Bilgi Sistemi Uygulaması
The Development Of Web Based Applications Creating User Sessions With Isapi
This list is updated automatically from ORCID.