TR-Dizin İndeksli Yayınlar Koleksiyonu / TR Dizin Indexed Publications Collection
Permanent URI for this collectionhttps://hdl.handle.net/20.500.11779/1927
Browse
4 results
Search Results
Article Mention Detection in Turkish Coreference Resolution(Tubitak Scientific & Technological Research Council Turkey, 2024) Demir, Seniz; Akdag, Hanifi IbrahimA crucial step in understanding natural language is detecting mentions that refer to real-world entities in a text and correctly identifying their boundaries. Mention detection is commonly considered a preprocessing step in coreference resolution which is shown to be helpful in several language processing applications such as machine translation and text summarization. Despite recent efforts on Turkish coreference resolution, no standalone neural solution to mention detection has been proposed yet. In this article, we present two models designed for detecting Turkish mentions by using feed-forward neural networks. Both models extract all spans up to a fixed length from input text as candidates and classify them as mentions or not mentions. The models differ in terms of how candidate text spans are represented. The first model represents a span by focusing on its first and last words, whereas the representation also covers the preceding and proceeding words of a span in the second model. Mention span representations are formed by using contextual embeddings, part-of-speech embeddings, and named-entity embeddings of words in interest where contextual embeddings are obtained from pretrained Turkish language models. In our evaluation studies, we not only assess the impact of mention representation strategies on system performance but also demonstrate the usability of different pretrained language models in resolution task. We argue that our work provides useful insights to the existing literature and the first step in understanding the effectiveness of neural architectures in Turkish mention detection.Article Citation - WoS: 1Citation - Scopus: 2Determining and Evaluating New Store Locations Using Remote Sensing and Machine Learning(Tübitak, 2021) Ünsalan, Cem; Turgay, Zeynep Zerrin; Küçükaydın, Hande; Höke, BerkanDecision making for store locations is crucial for retail companies as the profit depends on the location. The key point for correct store location is profit approximation, which is highly dependent on population of the corresponding region, and hence, the volume of the residential area. Thus, estimating building volumes provides insight about the revenue if a new store is about to be opened there. Remote sensing through stereo/tri-stereo satellite images provides wide area coverage as well as adequate resolution for three dimensional reconstruction for volume estimation. We reconstruct 3D map of corresponding region with the help of semiglobal matching and mask R-CNN algorithms for this purpose. Using the existing store data, we construct models for estimating the revenue based on surrounding building volumes. In order to choose the right location, the suitable utility model, which calculates store revenues, shouldbe rigorously determined. Moreover, model parameters should be assessed as correctly as possible. Instead of using randomly generated parameters, we employ remote sensing, computer vision, and machine learning techniques, which provide a novel way for evaluating new store locations.Article Citation - WoS: 4Citation - Scopus: 6Consumer Loans' First Payment Default (fpd) Detection and Predictive Model(TUBITAK SCIENTIFIC & TECHNICAL RESEARCH COUNCIL, 2020) Sevgili, Türkan; Koç, Utku; Koç, UtkuThe project is based on the opinion that whether the loan applications which are profitable could be granted instead of prone the default (FPD) ones by using predictive models in machine learning by the credit decision authorities in banking sector. Default Loan (also called non-performing loan) occurs when there is a failure to meet bank conditions and cannot be repaid in accordance with the terms of the loan which has reached its maturity. This report is a research effort in the analysis of default loan applicants, especially FPD, from a real dataset obtained from a bank. Expectation from the study is that increase the efficiency of consumer loan allocation by providing predictive analysis of the consumer behavior concerning loan’s first payment default. FPD detection analysis is a crucial role for the determination of consumer loans at the application level. The study also provides an understanding on the reasons of non-performing loans and helps to manage credit risks more consciously. The methods proposed in this study can be extended to other individual consumer loans such as car credits and mortgage.Article Citation - WoS: 8Citation - Scopus: 8Zaman Pencereli ve Değişken Başlama Zamanlı Bir Araç Rotalama Problemi için Sütun Türetme Temelli Matsezgiseller(DergiPark, 2019) Küçükaydın, HandeIn this study, a vehicle routing problem with time windows is investigated, where the costs depend on the total duration of vehicle routes and the starting time from the depot for each vehicle is determined by a decision maker. In order to solve the problem, two column generation based mat-heuristics are developed, where the first one makes use of the iterated local search and the second one uses the variable neighbourhood search. In order to assess the accuracy of the mat-heuristics, they are first compared with an exact algorithm on small instances taken from the literature. Since their performance are quite satisfactory, they are further tested on 87 large instances by running each algorithm 3 times for each instance. The computational results prove that the mat-heuristic using the variable neighbourhood search outperforms the other one. Hence, this enables to obtain a good feasible solution in a very short time when it is not possible to solve large instances with an exact solution method in a reasonable CPU time.