This paper presents a quantitative performance analysis of two different approaches to the lemmat... more This paper presents a quantitative performance analysis of two different approaches to the lemmatization of the Czech text data. The first one is based on manually prepared dictionary of lemmas and set of derivation rules while the second one is based on automatic inference of the dictionary and the rules from training data. The comparison is done by evaluating the mean Generalized Average Precision (mGAP) measure of the lemmatized documents and search queries in the set of information retrieval (IR) experiments. Such method is suitable for efficient and rather reliable comparison of the lemmatization performance since a correct lemmatization has proven to be crucial for IR effectiveness in highly inflected languages. Moreover, the proposed indirect comparison of the lemmatizers circumvents the need for manually lemmatized test data which are hard to obtain and also face the problem of incompatible sets of lemmas across different systems. ⋆
This paper presents the details of participation of DEMIR (Dokuz Eylül University Multimedia Info... more This paper presents the details of participation of DEMIR (Dokuz Eylül University Multimedia Information Retrieval) research team to the Share/CLEF eHealth 2014. This year, we participated to task 3a: monolingual user-centered health information retrieval. In this task, we focused to apply query expansion techniques selectively to some queries to improve the performance of information retrieval. Thus, we first extracted some statistical features from queries such as length of query, sum and intersect of document frequencies of each query term etc. We develop a system to predict if a query is to be expanded or not. Then, we trained our system with previous year’s data. Then, we applied a query expansion method only to the queries, which are selected by the system. The results show that the approach we proposed slightly improves our baseline retrieval performance in terms of P@10.
2007 IEEE 15th Signal Processing and Communications Applications, 2007
Abstract In this study, we have presented a speech vs nonspeech segmentation of audio signals ext... more Abstract In this study, we have presented a speech vs nonspeech segmentation of audio signals extracted from video. We have used 4330 seconds of audio signal extracted from" Lost" TV series for training. Our training set is automatically builded by using timestamp information exists in subtitles. After that, silence areas within those speech areas are discarded with a further study. Then, standard deviation of MFCC feature vectors of size 20 have been obtained. Finally, Support Vector Machines (SVM) is used with one-vs-all ...
2006 IEEE 14th Signal Processing and Communications Applications, 2006
Abstract This study presents a new method to recognize facial emotional expression which is all a... more Abstract This study presents a new method to recognize facial emotional expression which is all accepted globally that has a great role in human communication. The method uses basic image processing techniques and based on curve fitting on mouth region and able to detect happiness, surprise and sadness emotions within the universally accepted emotional expressions (angry, disgust, fear, happy, sad and surprise). The proposed approach is tested on a test bed containing total of 78 human face images of 13 different people with a ...
2007 IEEE 15th Signal Processing and Communications Applications, 2007
ABSTRACT In this study, we examine the effects of higher order statistics of timbral features to ... more ABSTRACT In this study, we examine the effects of higher order statistics of timbral features to improve performance of genre classification. It was seen that the first and second order statistics of the features extracted, in this research, is not as discriminative as the third and forth order statistics of the features. For the purpose of designing a classifier, which could be used for real time applications in future studies, randomly taken 3 second-long segments are used for classification. Out of 225 songs from 3 genres, 180 of them are used for training and 45 of them are used for testing. Five different lists that are created using different train and test sets are used to reduce the dependency of the results to the test set while increasing the number of validation data. Average values of validation test results are compared with the results of the similar works, which are based on MIDI format, using the same data set.
This study is based on indexing music sequences. In this respect, we consider representation of m... more This study is based on indexing music sequences. In this respect, we consider representation of music information and indexing music sequences with suffix trees. Current suffix tree algorithms gen- erally aim at indexing biological sequences such as DNA. Nevertheless, the algorithms do not consider the properties of music sequences. In order to fulfill such leak, we modify the physical structure of the suffix trees. Moreover, we present that clustering the music sequences and indexing each cluster by a different suffix tree enhance the performance. We have tested our approaches on MIDI music files and made comparisons with early approaches. Performance experiments show that our approach outper- forms the earlier approaches in the literature.
This paper presents an ontology-based annotation system and BI-RADS (Breast Imaging Reporting and... more This paper presents an ontology-based annotation system and BI-RADS (Breast Imaging Reporting and Data System) score reasoning with Semantic Web technologies in mammography. The annotation system is based on the Mammography Annotation Ontology (MAO) where the BI-RADS score reasoning works. However, ontologies are based on crisp logic and they cannot handle uncertainty. Consequently, we propose a Bayesian-based approach to model uncertainty in mammography ontology and make reasoning possible using BI-RADS scores with SQWRL (Semantic Query-enhanced Web Rule Language). First, we give general information about our system and present details of mammography annotation ontology, its main concepts and relationships. Then, we express uncertainty in mammography and present approaches to handle uncertainty issues. System is evaluated with a manually annotated dataset DEMS (Dokuz Eylul University Mammography Set) and DDSM (Digital Database for Screening Mammography). We give the result of experimentations in terms of accuracy, sensitivity, precision and uncertainty level measures.
Computer Methods and Programs in Biomedicine, 2013
c o m p u t e r m e t h o d s a n d p r o g r a m s i n b i o m e d i c i n e 1 1 0 ( 2 0 1 3 ) 1... more c o m p u t e r m e t h o d s a n d p r o g r a m s i n b i o m e d i c i n e 1 1 0 ( 2 0 1 3 ) 150-159 a b s t r a c t Many computer aided diagnosis (CAD) systems help radiologist on difficult task of mass detection in a breast mammogram and, besides, they also provide interpretation about detected mass. One of the most crucial information of a mass is its shape and contour, since it provides valuable information about spread ability of a mass. However, accuracy of shape recognition of a mass highly related with the precision of detected mass contours.
This paper presents a new approach to musical genre classification by using randomly selected max... more This paper presents a new approach to musical genre classification by using randomly selected maximum 3000 ms long segment of each recording. Recordings are classified into three root genres as jazz, classical and pop. A k-Nearest Neighbor (k-NN) classifier with pitch, harmony and rhythm features, independent from instruments, and a training set of 180 MIDI recordings is used for classification. The classifier is evaluated both quantitatively and cognitively using 45 MIDI recordings from outside the training set and considered for specificity, selectivity and accuracy measures. Experiments have demonstrated the good accuracy of the classifier and its high potential of use.
In this study we present an intelligent system proposal to help on assigning ICD10 codes to medic... more In this study we present an intelligent system proposal to help on assigning ICD10 codes to medical records. System is based on classification of free-form texts fields for a given medical record, and uses Terrier information retrieval engine to handle unstructured text data. In this way, we propose a new effective and efficient text classification method, which is based on information retrieval techniques. To evaluate the new approach we compiled a new Turkish medical collection, which includes approximately 50K preprocessed and anonimized medical records taken from Ankara University Hospital information systems. We measured the performance of proposed system in terms accuracy, selectivity and sensitivity. Özet: Bu çalışmada tıbbi kayıtlara ICD kodlarının atanmasına yardımcı olacak akıllı bir sistemin önerisi sunulmuştur. Sistem temel olarak verilen bir sağlık kaydında yer alan serbest formda yazılmış metin bilgilerinin sınıflandırmasına dayalıdır ve serbest formda yazılmış metin sahalarını işlemek için açık kaynak kodlu Terrier bilgi gerigetirim sistemini kullanır. Bu anlamda çalışmayı Türkçe tıbbı dokümanların sınıflandırması olarak da değerlendirmek mümkündür. Bu bağlamda, serbest formda yazılmış metinlerin sınıflandırılmasında bilgi gerigetirim tekniklerini kullanan etkin ve verimli bir sınıflandırma yaklaşımı öneriyoruz. Sistemin sınanması için Ankara Üniversitesi Hastanesi bilgi sisteminden yaklaşık 50 bin kayıt alınarak, önişlemden geçirip anonimize ederek, bir Türkçe tıbbi veri seti oluşturduk. Önerdiğimiz sistemin başarım performansları doğruluk, özgüllük ve duyarlılık ölçeklerine göre ölçülmüştür.
Rapidly expanding image collections in radiology departments create an urgent need to develop new... more Rapidly expanding image collections in radiology departments create an urgent need to develop new tools to manage these visual databases. This study aims at creating ontology based Case Based Retrieval model for mammographic examinations by using SWRL and SQWRL. This paper contains, explanation of mammography ontology, annotation tool, dataset and sample query results. Özet: Hastanelerin Radyoloji bölümlerinde üretilen görüntü koleksiyonlarının çok hızlı bir şekilde artması, bu görsel veri tabanlarını yönetmek için yeni araçlar geliştirilmesi ihtiyacını doğurmuştur. Bu çalışmanın amacı, anlamsal web teknolojilerinden SWRL ve SQWRL kullanarak mamografik incelemeler için ontoloji tabanlı bir Vaka Tabanlı Sorgulama sisteminin geliştirilmesidir. Bu amaçla oluşturulan mamografi ontolojisi, betimleme uygulaması, veri kümesi ve örnek sorgu sonuçları makale içersinde açıklanmıştır.
In this work, we present a reference dataset for human and machine learning in mammographic resea... more In this work, we present a reference dataset for human and machine learning in mammographic researches. The Dataset is called MuDMaDs, for "a Multipurpose Digital Mammography Reference Dataset". To accomplish this work, an ontology that includes all the details of mammography is created at first. Then, the candidate cases to be in the dataset are automatically selected by a boolean information retrieval system retrospectively from the cases that have been examined from 2004 to 2008 in the department of Radiodiagnostics at Dokuz Eylül University Hospital. In the final step, the mammograms of these cases are examined by the domain experts to select the cases to be in the dataset and the necessary corrections and annotations over them are done by using the application developed.
Đçerik Tabanlı Görüntü Erişim (ĐTGE) sistemleri büyük görüntü arşivlerinde, dışsal özellikler yan... more Đçerik Tabanlı Görüntü Erişim (ĐTGE) sistemleri büyük görüntü arşivlerinde, dışsal özellikler yanında görüntülerden otomatik olarak ayıklanan görsel alt düzey öznitelikler aracılığı ile erişilmesine olanak sağlarlar. Bu çalışmanın amacı, alt düzey görsel tanımlayacılar ile üst düzey kavramların birlikte kullanımına olanak verecek bir ĐTGE modeli için, bazı alt düzey özniteliklerin çıkartılmasıdır. Bu çalışmanın katkısı, anılan ĐTGE modelinde kullanılmak üzere çeşitli özniteliklerin modele katkılarının incelenerek en uygun özniteliklerin belirlenmesi için bir alt yapı oluşturmaktır. Çalışmada temel olarak: meme görüntülerinin bölütlenmesi, görüntü veritabanının hazırlanması, bölütlenmiş meme görüntülerinden bölgesel özniteliklerin çıkarılması ve çıkarılan özniteliklerin öğrenme metotları kullanılarak meme tipinin kestirimi gerçekleştirilmiştir.
This paper presents a quantitative performance analysis of two different approaches to the lemmat... more This paper presents a quantitative performance analysis of two different approaches to the lemmatization of the Czech text data. The first one is based on manually prepared dictionary of lemmas and set of derivation rules while the second one is based on automatic inference of the dictionary and the rules from training data. The comparison is done by evaluating the mean Generalized Average Precision (mGAP) measure of the lemmatized documents and search queries in the set of information retrieval (IR) experiments. Such method is suitable for efficient and rather reliable comparison of the lemmatization performance since a correct lemmatization has proven to be crucial for IR effectiveness in highly inflected languages. Moreover, the proposed indirect comparison of the lemmatizers circumvents the need for manually lemmatized test data which are hard to obtain and also face the problem of incompatible sets of lemmas across different systems. ⋆
This paper presents the details of participation of DEMIR (Dokuz Eylül University Multimedia Info... more This paper presents the details of participation of DEMIR (Dokuz Eylül University Multimedia Information Retrieval) research team to the Share/CLEF eHealth 2014. This year, we participated to task 3a: monolingual user-centered health information retrieval. In this task, we focused to apply query expansion techniques selectively to some queries to improve the performance of information retrieval. Thus, we first extracted some statistical features from queries such as length of query, sum and intersect of document frequencies of each query term etc. We develop a system to predict if a query is to be expanded or not. Then, we trained our system with previous year’s data. Then, we applied a query expansion method only to the queries, which are selected by the system. The results show that the approach we proposed slightly improves our baseline retrieval performance in terms of P@10.
2007 IEEE 15th Signal Processing and Communications Applications, 2007
Abstract In this study, we have presented a speech vs nonspeech segmentation of audio signals ext... more Abstract In this study, we have presented a speech vs nonspeech segmentation of audio signals extracted from video. We have used 4330 seconds of audio signal extracted from" Lost" TV series for training. Our training set is automatically builded by using timestamp information exists in subtitles. After that, silence areas within those speech areas are discarded with a further study. Then, standard deviation of MFCC feature vectors of size 20 have been obtained. Finally, Support Vector Machines (SVM) is used with one-vs-all ...
2006 IEEE 14th Signal Processing and Communications Applications, 2006
Abstract This study presents a new method to recognize facial emotional expression which is all a... more Abstract This study presents a new method to recognize facial emotional expression which is all accepted globally that has a great role in human communication. The method uses basic image processing techniques and based on curve fitting on mouth region and able to detect happiness, surprise and sadness emotions within the universally accepted emotional expressions (angry, disgust, fear, happy, sad and surprise). The proposed approach is tested on a test bed containing total of 78 human face images of 13 different people with a ...
2007 IEEE 15th Signal Processing and Communications Applications, 2007
ABSTRACT In this study, we examine the effects of higher order statistics of timbral features to ... more ABSTRACT In this study, we examine the effects of higher order statistics of timbral features to improve performance of genre classification. It was seen that the first and second order statistics of the features extracted, in this research, is not as discriminative as the third and forth order statistics of the features. For the purpose of designing a classifier, which could be used for real time applications in future studies, randomly taken 3 second-long segments are used for classification. Out of 225 songs from 3 genres, 180 of them are used for training and 45 of them are used for testing. Five different lists that are created using different train and test sets are used to reduce the dependency of the results to the test set while increasing the number of validation data. Average values of validation test results are compared with the results of the similar works, which are based on MIDI format, using the same data set.
This study is based on indexing music sequences. In this respect, we consider representation of m... more This study is based on indexing music sequences. In this respect, we consider representation of music information and indexing music sequences with suffix trees. Current suffix tree algorithms gen- erally aim at indexing biological sequences such as DNA. Nevertheless, the algorithms do not consider the properties of music sequences. In order to fulfill such leak, we modify the physical structure of the suffix trees. Moreover, we present that clustering the music sequences and indexing each cluster by a different suffix tree enhance the performance. We have tested our approaches on MIDI music files and made comparisons with early approaches. Performance experiments show that our approach outper- forms the earlier approaches in the literature.
This paper presents an ontology-based annotation system and BI-RADS (Breast Imaging Reporting and... more This paper presents an ontology-based annotation system and BI-RADS (Breast Imaging Reporting and Data System) score reasoning with Semantic Web technologies in mammography. The annotation system is based on the Mammography Annotation Ontology (MAO) where the BI-RADS score reasoning works. However, ontologies are based on crisp logic and they cannot handle uncertainty. Consequently, we propose a Bayesian-based approach to model uncertainty in mammography ontology and make reasoning possible using BI-RADS scores with SQWRL (Semantic Query-enhanced Web Rule Language). First, we give general information about our system and present details of mammography annotation ontology, its main concepts and relationships. Then, we express uncertainty in mammography and present approaches to handle uncertainty issues. System is evaluated with a manually annotated dataset DEMS (Dokuz Eylul University Mammography Set) and DDSM (Digital Database for Screening Mammography). We give the result of experimentations in terms of accuracy, sensitivity, precision and uncertainty level measures.
Computer Methods and Programs in Biomedicine, 2013
c o m p u t e r m e t h o d s a n d p r o g r a m s i n b i o m e d i c i n e 1 1 0 ( 2 0 1 3 ) 1... more c o m p u t e r m e t h o d s a n d p r o g r a m s i n b i o m e d i c i n e 1 1 0 ( 2 0 1 3 ) 150-159 a b s t r a c t Many computer aided diagnosis (CAD) systems help radiologist on difficult task of mass detection in a breast mammogram and, besides, they also provide interpretation about detected mass. One of the most crucial information of a mass is its shape and contour, since it provides valuable information about spread ability of a mass. However, accuracy of shape recognition of a mass highly related with the precision of detected mass contours.
This paper presents a new approach to musical genre classification by using randomly selected max... more This paper presents a new approach to musical genre classification by using randomly selected maximum 3000 ms long segment of each recording. Recordings are classified into three root genres as jazz, classical and pop. A k-Nearest Neighbor (k-NN) classifier with pitch, harmony and rhythm features, independent from instruments, and a training set of 180 MIDI recordings is used for classification. The classifier is evaluated both quantitatively and cognitively using 45 MIDI recordings from outside the training set and considered for specificity, selectivity and accuracy measures. Experiments have demonstrated the good accuracy of the classifier and its high potential of use.
In this study we present an intelligent system proposal to help on assigning ICD10 codes to medic... more In this study we present an intelligent system proposal to help on assigning ICD10 codes to medical records. System is based on classification of free-form texts fields for a given medical record, and uses Terrier information retrieval engine to handle unstructured text data. In this way, we propose a new effective and efficient text classification method, which is based on information retrieval techniques. To evaluate the new approach we compiled a new Turkish medical collection, which includes approximately 50K preprocessed and anonimized medical records taken from Ankara University Hospital information systems. We measured the performance of proposed system in terms accuracy, selectivity and sensitivity. Özet: Bu çalışmada tıbbi kayıtlara ICD kodlarının atanmasına yardımcı olacak akıllı bir sistemin önerisi sunulmuştur. Sistem temel olarak verilen bir sağlık kaydında yer alan serbest formda yazılmış metin bilgilerinin sınıflandırmasına dayalıdır ve serbest formda yazılmış metin sahalarını işlemek için açık kaynak kodlu Terrier bilgi gerigetirim sistemini kullanır. Bu anlamda çalışmayı Türkçe tıbbı dokümanların sınıflandırması olarak da değerlendirmek mümkündür. Bu bağlamda, serbest formda yazılmış metinlerin sınıflandırılmasında bilgi gerigetirim tekniklerini kullanan etkin ve verimli bir sınıflandırma yaklaşımı öneriyoruz. Sistemin sınanması için Ankara Üniversitesi Hastanesi bilgi sisteminden yaklaşık 50 bin kayıt alınarak, önişlemden geçirip anonimize ederek, bir Türkçe tıbbi veri seti oluşturduk. Önerdiğimiz sistemin başarım performansları doğruluk, özgüllük ve duyarlılık ölçeklerine göre ölçülmüştür.
Rapidly expanding image collections in radiology departments create an urgent need to develop new... more Rapidly expanding image collections in radiology departments create an urgent need to develop new tools to manage these visual databases. This study aims at creating ontology based Case Based Retrieval model for mammographic examinations by using SWRL and SQWRL. This paper contains, explanation of mammography ontology, annotation tool, dataset and sample query results. Özet: Hastanelerin Radyoloji bölümlerinde üretilen görüntü koleksiyonlarının çok hızlı bir şekilde artması, bu görsel veri tabanlarını yönetmek için yeni araçlar geliştirilmesi ihtiyacını doğurmuştur. Bu çalışmanın amacı, anlamsal web teknolojilerinden SWRL ve SQWRL kullanarak mamografik incelemeler için ontoloji tabanlı bir Vaka Tabanlı Sorgulama sisteminin geliştirilmesidir. Bu amaçla oluşturulan mamografi ontolojisi, betimleme uygulaması, veri kümesi ve örnek sorgu sonuçları makale içersinde açıklanmıştır.
In this work, we present a reference dataset for human and machine learning in mammographic resea... more In this work, we present a reference dataset for human and machine learning in mammographic researches. The Dataset is called MuDMaDs, for "a Multipurpose Digital Mammography Reference Dataset". To accomplish this work, an ontology that includes all the details of mammography is created at first. Then, the candidate cases to be in the dataset are automatically selected by a boolean information retrieval system retrospectively from the cases that have been examined from 2004 to 2008 in the department of Radiodiagnostics at Dokuz Eylül University Hospital. In the final step, the mammograms of these cases are examined by the domain experts to select the cases to be in the dataset and the necessary corrections and annotations over them are done by using the application developed.
Đçerik Tabanlı Görüntü Erişim (ĐTGE) sistemleri büyük görüntü arşivlerinde, dışsal özellikler yan... more Đçerik Tabanlı Görüntü Erişim (ĐTGE) sistemleri büyük görüntü arşivlerinde, dışsal özellikler yanında görüntülerden otomatik olarak ayıklanan görsel alt düzey öznitelikler aracılığı ile erişilmesine olanak sağlarlar. Bu çalışmanın amacı, alt düzey görsel tanımlayacılar ile üst düzey kavramların birlikte kullanımına olanak verecek bir ĐTGE modeli için, bazı alt düzey özniteliklerin çıkartılmasıdır. Bu çalışmanın katkısı, anılan ĐTGE modelinde kullanılmak üzere çeşitli özniteliklerin modele katkılarının incelenerek en uygun özniteliklerin belirlenmesi için bir alt yapı oluşturmaktır. Çalışmada temel olarak: meme görüntülerinin bölütlenmesi, görüntü veritabanının hazırlanması, bölütlenmiş meme görüntülerinden bölgesel özniteliklerin çıkarılması ve çıkarılan özniteliklerin öğrenme metotları kullanılarak meme tipinin kestirimi gerçekleştirilmiştir.
Uploads
Papers by Adil Alpkocak