5 sonuçlar
Arama Sonuçları
Listeleniyor 1 - 5 / 5
Yayın A FST description of noun and verb morphology of Azarbaijani Turkish(Association for Computational Linguistics (ACL), 2021) Ehsani, Razieh; Özenç, Berke; Solak, Ercan; Drewes F.We give a FST description of nominal and finite verb morphology of Azarbaijani Turkish. We use a hybrid approach where nominal inflection is expressed as a slot-based paradigm and major parts of verb inflection are expressed as optional paths on the FST. We collapse adjective and noun categories in a single nominal category as they behave similarly as far as their paradigms are concerned. Thus, we defer a more precise identification of POS to further down the NLP pipeline.Yayın Co-training using prosodic, lexical and morphological information for automatic sentence segmentation of Turkish spoken language(Işık Üniversitesi, 2018-01-15) Dalva, Doğan; Güz, Ümit; Işık Üniversitesi, Fen Bilimleri Enstitüsü, Elektronik Mühendisliği Doktora ProgramıSentence segmentation of speech aims detecting sentence boundaries in a stream of words output by the speech recognizer. Sentence segmentation is a preliminary step toward speech understanding. It is of particular importance for speech related applications, as most of the further processing steps; such as parsing, machine translation and information extraction, assume the presence of sentence boundaries. Typically, statistical methods require a huge amount of manually labeled data, which is time and labor consuming process to prepare. In this work, novel multiview semi-supervised learning strategies for the solution of sentence segmentation problem are proposed. The aim of this work is to and effective semi-supervised machine learning strategies when only a small set of sentence boundary labeled data is available. This work proposes three-view co-training and committee-based strategies incorporating with agreement, disagreement and self-combined strategies using lexical, morphological and prosodic information, and investigates performance of the proposed learning strategies against baseline, self-training and co-training. The experimental results show that the proposed learning strategies highly improve the sentence segmentation problem, since data sets can be represented by three redundantly suffcient and disjoint feature sets.Yayın Morphological analyser for Turkish(Işık Üniversitesi, 2018-01-25) Özenç, Berke; Solak, Ercan; Işık Üniversitesi, Fen Bilimleri Enstitüsü, Bilgisayar Mühendisliği Yüksek Lisans ProgramıNatural Language Processing is one one the fields of work in computer science and specializes in text summarization, machine translation and many various topics. Morphology is one of the Natural Language Processing features which analyses the words with its suxes. A words meaning can change according to the sux that it takes. Turkish is an agglutinative language with rich morphological structure and set of suxes. This features of Turkish result in complex morphology structure. In this study, we present an analyser for Modern Anatolian Turkish which has high coverage on suffixes and morphological rules of Turkish. Two-Level transformation method which is convenient to design morphology of a language, consists our base of approach. We used HFST which is a Finite State Transducer implementation, as our implementation technique. The analyser covers all morphological and phonetic rules that exist in Turkish and contains a lexicon which consist of today's Turkish words. The analyser is publicly available and can be used on http://ddil.isikun.edu.tr/mortur.Yayın A novel approach to non-invasive intracranial pressure wave monitoring: a pilot healthy brain study(Multidisciplinary Digital Publishing Institute (MDPI), 2025-06-28) Karaliunas, Andrius; Bartusis, Laimonas; Krakauskaite, Solventa; Chaleckas, Edvinas; Deimantavicius, Mantas; Hamarat, Yasin; Petkus, Vytautas; Stulge, Toma; Ratkunas, Vytenis; Çelikkaya, Güven; Januleviciene, Ingrida; Ragauskas, ArminasIntracranial pressure (ICP) pulse wave morphology, including the ratios of the three characteristic peaks (P1, P2, and P3), offers valuable insights into intracranial dynamics and brain compliance. Traditional invasive methods for ICP pulse wave monitoring pose significant risks, highlighting the need for non-invasive alternatives. This pilot study investigates a novel non-invasive method for monitoring ICP pulse waves through closed eyelids, using a specially designed, liquid-filled, fully passive sensor system named ‘Archimedes 02’. To our knowledge, this is the first technological approach that enables the non-invasive monitoring of ICP pulse waveforms via closed eyelids. This study involved 10 healthy volunteers, aged 26–39 years, who underwent resting-state non-invasive ICP pulse wave monitoring sessions using the ‘Archimedes 02’ device while in the supine position. The recorded signals were processed to extract pulse waves and evaluate their morphological characteristics. The results indicated successful detection of pressure pulse waves, showing the expected three peaks (P1, P2, and P3) in all subjects. The calculated P2/P1 ratios were 0.762 (SD = ±0.229) for the left eye and 0.808 (SD = ±0.310) for the right eye, suggesting normal intracranial compliance across the cohort, despite variations observed in some individuals. Physiological tests—the Valsalva maneuver and the Queckenstedt test, both performed in the supine position—induced statistically significant increases in the P2/P1 and P3/P1 ratios, supporting the notion that non-invasively recorded pressure pulse waves, measured through closed eyelids, reflect intracranial volume and pressure dynamics. Additionally, a transient hypoemic/hyperemic response test performed in the upright position induced signal changes in pressure recordings from the ‘Archimedes 02’ sensor that were consistent with intact cerebral blood flow autoregulation, aligning with established physiological principles. These findings indicate that ICP pulse waves and their dynamic changes can be monitored non-invasively through closed eyelids, offering a potential method for brain monitoring in patients for whom invasive procedures are not feasible.Yayın Türkçe için biçimbirim temelli bir bileşen grameri yaklaşımı(Beykoz Üniversitesi, 2024-12-26) Özenç, Berke; Solak, ErcanDilin modellenmesi, dil çalışmalarında önemli bir temel olarak yer alır. Farklı modelleme yöntemleri, farklı diller için uyarlanabilir olsa da bu uyarlamalar, hedef dil için her zaman yeterli olmayabilir. Bu durumdan en çok biçimbirimsel açıdan zengin diller etkilenir. Böyle bir dil için hazırlanacak model kurgulanırken dilin evrensel olarak ortak olan özelliklerinin yanı sıra, dilin kendine özgü özelliklerine odaklanılmalıdır. Bu makalede, bağımlı biçimbirim bakımından zengin bir görünüm sunan Türkçe ele alınarak uyarlanan gramer sunulmuştur. Çalışmada açıklanan gramer temelleri geleneksel üretici gramer yönteminden uyarlanmıştır. Bununla birlikte, sunulan gramer, biçimbirimleri söz dizimi elemanı olarak geleneksel söz dizimi elemanlarıyla birlikte, söz dizimine olan etkilerini ele almasıyla ve kullanılan özel bileşen kümesiyle geleneksel üretici gramer yöntemden ayrılır. Geleneksel yöntemden farklı olarak önerilen gramerde, tümce çözümlemesine sözcüklerden değil, biçimbirim elemanları olan sözcük gövdeleri, ekler, biçimbirimler ve bu gibi elemanların oluşturduğu gruplardan başlanır. Buna ek olarak Türkçenin söz dizimsel ve birimbirimsel özelliklerine göre kurgulanan bir bileşen kümesi de sunulmuştur. Sunulan bileşen kümesi, tümce, ad öbeği, eylem öbeği, belirteç öbeği gibi geleneksel sözdizimsel bileşenleri, öbek gövdesi olarak adlandırılan ara bir yapıyı ve çoğul eki, durum eki, zaman çekimi eki gibi, biçimbirimleri veya biçimbirim gruplarını temsil eden bileşenleri içerir.
