Senem Tanberk

Senem Tanberk

Auto-generated profile, based on publicly available information

Senem Tanberk

Huawei Trkiye Research and Development Center,Research and Innovation,stanbul,Trkiye

Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
21 January 2020 in IEEE Access
Verified
Journal Article
Open access
STSenem Tanberk
Senem Tanberk
Zeynep Hilal KİLİMCİZeynep Hilal KİLİMCİ
Zeynep Hilal KİLİMCİ
DTDilek Bilgin Tukel
Dilek Bilgin Tukel
MUMitat Uysal
Mitat Uysal
SASelim Akyokus
Selim Akyokus
Human activity recognition is a challenging problem with many applications including visual surveillance, human-computer interactions, autonomous driving and entertainment. In this study, we propose a hybrid deep model to understand and interpret videos focusing on human activity recognition. The proposed architecture is constructed combining dense optical flow approach and auxiliary movement information in video datasets using deep learning methodologies. To the best of our knowledge, this is the first study based on a novel combination of 3D-convolutional neural networks (3D-CNNs) fed by optical flow and long short-term memory networks (LSTM) fed by auxiliary information over video frames for the purpose of human activity recognition. The contributions of this paper are sixfold. First, a 3D-CNN, also called multiple frames is employed to determine the motion vectors. With the same purpose, the 3D-CNN is secondly used for dense optical flow, which is the distribution of apparent velocities of movement in captured imagery data in video frames. Third, the LSTM is employed as auxiliary information in video to recognize hand-tracking and objects. Fourth, the support vector machine algorithm is utilized for the task of classification of videos. Fifth, a wide range of comparative experiments are conducted on two newly generated chess datasets, namely the magnetic wall chess board video dataset (MCDS), and standard chess board video dataset (CDS) to demonstrate the contributions of the proposed study. Finally, the experimental results reveal that the proposed hybrid deep model exhibits remarkable performance compared to the state-of-the-art studies.

21 January 2021
Verified
Conference Paper
Closed access
STSenem Tanberk
Senem Tanberk
DTDilek Bilgin Tukel
Dilek Bilgin Tukel
The recent advance in artificial neural networks is an inspiration for automatic music generation. Deep learning algorithms help to produce pleasing melodies. They lead the creativity of musicians to be reproduced in digital environments. The proposed system learns from the Turkish popular music and then produces new music. In this study, our goal is to generate melody with a specific style, such as unforgettable soundtracks admired widely. We proposed a novel combination of convolutional neural network (CNN) and long short-term memory (LSTM) network for music generation. The experimental results reveal that the proposed combined deep model exhibits remarkable music quality compared to the lstm-only deep model or cnn-only deep model. We also conducted a survey to evaluate the quality of the generated music. The survey results show that the introduced model is capable of producing better quality and more pleasant music compared to other state-of-the-art music generation methods.

9 June 2021
Verified
Conference Paper
Closed access
FTFerhat Tuncel
Ferhat Tuncel
BMBasri Mumcu
Basri Mumcu
STSenem Tanberk
Senem Tanberk
It is important to provide clinical guidance to patients according to their symptoms. Before going to the doctor, many people search for their symptoms via the internet and misdiagnose themselves. They usually think they have a much more serious illness than they have. Nowadays, clinical guidance tasks completed by internet searching, hospital information desk, or hear saying in society may cause latency in getting accurate health care. These results demand accurate and fast guidance to clinical services. The chatbot created based on a reliable database of symptoms can assist millions of patients to receive a more accurate preliminary diagnosis. However, a chatbot cannot substitute professional healthcare consulting and a visit to a doctor is a must. In this paper, we present an AI chatbot using a newly generated database for preliminary patient guidance and then show the results of the proposed architecture. The Chatbot for Preliminary Patient Guidance System (PGS) aims to guide patients to clinical services by checking symptoms automatically and to acquire required healthcare services quickly. Patients can get information about the possible disease (s) and related clinical service (s) by telling their symptoms to the PGS.

14 September 2022
Verified
Conference Paper
Closed access
STSenem Tanberk
Senem Tanberk
ODOnur Demir
Onur Demir
Data traffic prediction is essential for resource planning and allocation for service providers. Call Detail Records (CDR) provides invaluable information about user movements and behavior. However, the scale and complexity of CDR arise problems with its continuous usage in real-life issues. In this study, we propose a summary data structure out of CDR data to improve analysis performance. We then use this new data structure to make inferences using Multivariate Time Series analyses about the data traffic. We used several models, including Long Short-Term Memory networks (LSTM) and eXtreme Gradient Boosting (XGBoost), to verify the effectiveness of this approach. According to the results, our multivariate approach ensures usage trend capture. The research findings are efficient and suitable for predicting real-world network traffic based on usage type.

24 October 2024
Verified
Conference Paper
Closed access
BABerhan Türkü Ay
Berhan Türkü Ay
ÖMÖykü Berfin Mercan
Öykü Berfin Mercan
SHSelahattin Serdar Helli
Selahattin Serdar Helli
STSenem Tanberk
Senem Tanberk
In recent years, despite data limitations, sign language studies have attracted researchers' attention. For translating between spoken language and sign language gloss, which is considered a machine translation problem, data augmentation and state-of-the-art techniques have been proposed to overcome poor performance caused by data scarcity. This study focused on data augmentation for American Sign Language and British Sign Language using custom instructions ChatGPT4.0. Two different data selection methods were proposed to guide ChatGPT-4.0 in data generation. After data augmentation, the M2M100-418M model was fine-tuned on the original and augmented dataset. Experiments on the test data show that our method improves text-to-gloss translation performance.