I. Hakki Toroslu

Automatic reconstruction of broken 3-D surface objects

Abstract The problem of reconstruction of broken surface objects embedded in 3-D space is handled. A coordinate independent representation for the crack curves is developed. A new robust matching algorithm is proposed which serves for finding matching pieces even when some brittle pieces are missing. A prototype system having an X-based GUI has been developed. This system generates artifical wire–frame data of broken pieces (with some noise) for a pot-shaped 3-D object and then recombines it using the proposed algorithms.

A semantic-based user privacy protection framework for web services

Web service technology is an Internet-based distributed computing paradigm to address interoperability in heterogeneous distributed systems. In this paper, we present a privacy framework for Web services which allows user agents to automatically negotiate with Web services on the amount of personal information to be disclosed on behalf of the user. In developing this framework the following key privacy considerations are taken into account: revealing only the minimal pertinent information about the user, not to overwhelm the users while declaring their privacy preferences and requiring only limited user interaction. In the framework proposed, the Web services declare their input parameters as Mandatory or Optional and allow users to declare how much of their personal information can be made available to the services. The users specify their privacy preferences in different permission levels on the basis of a domain specific service ontology based on DAML-S. The major components of the system are a globally accessible context server which stores user preferences and a service registry where the services advertised and the service semantics are available.

Integrated search and alignment of protein structures

MOTIVATION Identification and comparison of similar three-dimensional (3D) protein structures has become an even greater challenge in the face of the rapidly growing structure databases. Here, we introduce Vorometric, a new method that provides efficient search and alignment of a query protein against a database of protein structures. Voronoi contacts of the protein residues are enriched with the secondary structure information and a metric substitution matrix is developed to allow efficient indexing. The contact hits obtained from a distance-based indexing method are extended to obtain high-scoring segment pairs, which are then used to generate structural alignments. RESULTS Vorometric is the first to address both search and alignment problems in the protein structure databases. The experimental results show that Vorometric is simultaneously effective in retrieving similar protein structures, producing high-quality structure alignments, and identifying cross-fold similarities. Vorometric outperforms current structure retrieval methods in search accuracy, while requiring com-parable running times. Furthermore, the structural superpositions produced are shown to have better quality and coverage, when compared with those of the popular structure alignment tools. AVAILABILITY Vorometric is available as a web service at http://bio.cse.ohio-state.edu/Vorometric

A genetic algorithm approach for verification of the syllable-based text compression technique

Provided that an easy mechanism exists for it, it is possible to decompose a text into strings that have lengths greater than one and occur frequently. Having in one hand the set of such frequently occurring strings and in the other the set of letters and symbols, it is possible to compress the text using Huffman coding over an alphabet which is a subset of the union of these two sets. Observations reveal that, in most cases, the maximal inclusion of the strings leads to an optimal length of compressed text. However, the verification of this prediction requires the consideration of all subsets in order to find the one that leads to the best compression. A genetic algorithm is devised and used for this search process. In Turkish texts, because of the agglutinative nature of the language, a highly regular syllable formation provides a useful testbed.

Transfer Learning Using Twitter Data for Improving Sentiment Classification of Turkish Political News

In this paper, we aim to determine the overall sentiment classification of Turkish political columns. That is, our goal is to determine whether the whole document has positive or negative opinion regardless of its subject. In order to enhance the performance of the classification, transfer learning is applied from unlabeled Twitter data to labeled political columns. A variation of self-taught learning has been proposed, and implemented for the classification. Different machine learning techniques, including support vector machine, maximum entropy classification, and Naive-Bayes has been used for the supervised learning phase. In our experiments we have obtained up to 26 % increase in the accuracy of the classification with the inclusion of the Twitter data into the sentiment classification of Turkish political columns using transfer learning.

Location prediction of mobile phone users using apriori-based sequence mining with multiple support thresholds

Due to the increasing use of mobile phones and their increasing capabilities, huge amount of usage and location data can be collected. Location prediction is an important task for mobile phone operators and smart city administrations to provide better services and recommendations. In this work, we propose a sequence mining based approach for location prediction of mobile phone users. More specifically, we present a modified Apriori-based sequence mining algorithm for the next location prediction, which involves use of multiple support thresholds for different levels of pattern generation process. The proposed algorithm involves a new support definition, as well. We have analyzed the behaviour of the algorithm under the change of threshold through experimental evaluation and the experiments indicate improvement in comparison to conventional Apriori-based algorithm.

Approximate Similarity Search in Genomic Sequence Databases Using Landmark-Guided Embedding

Similarity search in sequence databases is of paramount importance in bioinformatics research. As the size of the genomic databases increases, similarity search of proteins in these databases becomes a bottle-neck in large-scale studies, calling for more efficient methods of content-based retrieval. In this study, we present a metric-preserving, landmark-guided embedding approach to represent sequences in the vector domain in order to allow efficient indexing and similarity search. We analyze various properties of the embedding and show that the approximation achieved by the embedded representation is sufficient to achieve biologically relevant results. The approximate representation is shown to provide several orders of magnitude speed-up in similarity search compared to the exact representation, while maintaining comparable search accuracy.

Improving the prediction of page access by using semantically enhanced clustering

There are many parameters that may affect the navigation behaviour of web users. Prediction of the potential next page that may be visited by the web user is important, since this information can be used for prefetching or personalization of the page for that user. One of the successful methods for the determination of the next web page is to construct behaviour models of the users by clustering. The success of clustering is highly correlated with the similarity measure that is used for calculating the similarity among navigation sequences. This work proposes a new approach for determining the next web page by extending the standard clustering with the content-based semantic similarity method. Semantics of web-pages are represented as sets of concepts, and thus, user session are modelled as sequence of sets. As a result, session similarity is defined as an alignment of two sequences of sets. The success of the proposed method has been shown through applying it on real life web log data.

Partitioning and Scaling Signed Bipartite Graphs for Polarized Political Blogosphere

Blogosphere plays an increasingly important role as a forum for public debate. In this paper, given a mixed set of blogs debating a set of political issues from opposing camps, we use signed bipartite graphs for modeling debates, and we propose an algorithm for partitioning both the blogs, and the issues (i.e. topics, leaders, etc.) comprising the debate into binary opposing camps. Simultaneously, our algorithm scales both the blogs and the underlying issues on a univariate scale. Using this scale, a researcher can identify moderate and extreme blogs within each camp, and polarizing vs. unifying issues. Through performance evaluations we show that our proposed algorithm provides an effective solution to the problem, and performs much better than existing baseline algorithms adapted to solve this new problem. In our experiments, we used both real data from political blogosphere and US Congress records, as well as synthetic data which were obtained by varying polarization and degree distribution of the vertices of the graph to show the robustness of our algorithm.

Analyzing Implicit Aspects and Aspect Dependent Sentiment Polarity for Aspect-based Sentiment Analysis on Informal Turkish Texts

The web provides a suitable media for users to post comments on different topics. In most of such content, authors express different opinions on different features or aspects of the topic. In aspect based sentiment analysis, it is analyzed as to for which aspect which opinion is expressed. Once aspects are available, the next important step is to match aspects with correct sentiments. In this work, we investigate enhancements for two cases in matching step: extracting implicit aspects, and sentiment words whose polarity depends on the aspect. The techniques are applied on Turkish informal texts collected from a products forum. Experimental evaluation shows that additional steps applied for these cases improve the accuracy of aspect based sentiment analysis.