Aykut Koc Lab - Bilkent

Aykut Koc Lab is a research group at Bilkent University focusing on Natural Language Processing research.Some research on graph signal processing also going on.

[Aykut Koc Lab - Bilkent @KocLab_Bilkent]

3- by utilizing the parametric nature of FrFT that provides us to reach and select from infinitely many signal domains to represent time-series signals for better prediction.

Paper: https://t.co/0iNGio5rEk
Code: https://t.co/ujj5PUnW1h

[Aykut Koc Lab - Bilkent @KocLab_Bilkent]

1- Aykut Koc Lab has made a contribution at the intersection of classical signal processing and machine learning by introducing the fractional Fourier transform (FrFT) to the machine learning domain for time series analysis.-

[Aykut Koc Lab - Bilkent @KocLab_Bilkent]

2- Our architecture utilizing FrFT with Recurrent Neural Networks (RNNs) for time series prediction problem is now out at IEEE Signal Processing Letters! We show that FrFT-based RNN models give better prediction performances than those based on either time and frequency domain-

[Aykut Koc Lab - Bilkent @KocLab_Bilkent]

3- yet outperforms state-of-the-art imputation methods over benchmark datasets by 13.5-50.5%.

Paper: https://t.co/uRNTNASZXk
Code: https://t.co/thoWpSeZs6

[Aykut Koc Lab - Bilkent @KocLab_Bilkent]

1- NEWS (2022.11.25): Our work on the first transformer encoder-based missing value imputation is now out at IEEE Signal Processing Letters! We propose the Multivariate Time-Series Imputation with Transformers (MTSIT),-

[Aykut Koc Lab - Bilkent @KocLab_Bilkent]

2- a method that utilizes the transformer architecture in an unsupervised manner for missing value imputation. MTSIT uses only the encoder part of the transformer, which results in a remarkably efficient and small model with less than 6,000 parameters,-

[Aykut Koc Lab - Bilkent @KocLab_Bilkent]

2- to mix the tokens of the transformer.

Paper: https://t.co/DTGQh6Qay3

[Aykut Koc Lab - Bilkent @KocLab_Bilkent]

1- NEWS (2022.10.28): Aykut Koc Lab has made a fundamental contribution at the intersection of classical signal processing and machine learning by introducing an advanced and efficient transformer model (FrFNet) that utilizes the fractional Fourier transform (FrFT)-

[Aykut Koc Lab - Bilkent @KocLab_Bilkent]

2- This work presents a NER model for Turkish legal texts with a custom-made corpus as well as several NER architectures based on conditional random fields and bidirectional long-short-term memories (BiLSTMs) to address the task.

Paper: https://t.co/aY9Iox1D9d

[Aykut Koc Lab - Bilkent @KocLab_Bilkent]

1- NEWS (2022.07.11): KocLab has published the first legal domain-specific NER study in Turkish in the paper "Named Entity Recognition in Turkish Legal Texts" at Natural Language Engineering journal of Cambridge University Press.

[Aykut Koc Lab - Bilkent @KocLab_Bilkent]

Our group’s undergraduate member Mustafa Bozdağ will be continuing his graduate studies at Northeastern University starting from the next academic year. We wish him good luck and success…

[Aykut Koc Lab - Bilkent @KocLab_Bilkent]

Our group’s undergraduate member Enes Burak Bilgin will be continuing his graduate studies at the University of Southern California (USC) starting from the next academic year. We wish him good luck and success…

[Aykut Koc Lab - Bilkent @KocLab_Bilkent]

Our group’s undergraduate member Bünyamin Kartal will be continuing his graduate studies at the Massachusetts Institute of Technology (MIT) starting from the next academic year. We wish him good luck and success…

[Aykut Koc Lab - Bilkent @KocLab_Bilkent]

2- Cambridge University Press. This work brings the important bias issues to the attention of computational law community, where the importance of law for society has doubled the need for studying bias. Paper: https://t.co/CIttMCRBvN

[Aykut Koc Lab - Bilkent @KocLab_Bilkent]

1- KocLab has published the first study on detecting and eliminating gender, race and other societal bias issues present in word embeddings for the case of legal domain in the paper “Gender Bias in Legal Corpora and Debiasing It” at Natural Language Engineering journal of-

[Aykut Koc Lab - Bilkent @KocLab_Bilkent]

3- Paper: https://t.co/5BRC5sSkbz

[Aykut Koc Lab - Bilkent @KocLab_Bilkent]

1- In collaboration with Schütze Lab (LMU), Icon Lab (Bilkent) and ASELSAN Research Center, KocLab has published "Learning Interpretable Word Embeddings via Bidirectional Alignment of Dimensions with Semantic Concepts" in Information Processing and Management journal. The paper-

[Aykut Koc Lab - Bilkent @KocLab_Bilkent]

2- proposes the bidirectional imparting (BiImp) method which aligns both positive and negative dimension of the word embeddings with concepts in order to have more interpretable word embeddings. BiImp is also shown to be applied to mitigating gender bias in word embeddings.

[Aykut Koc Lab - Bilkent @KocLab_Bilkent]

2- The proposed method reached high rankings (including the first rank at one of the two sub-tasks) at the SemEval 2020 Task 1, which is the standardized evaluation framework of the field. Paper: https://t.co/rCP9llJlQF

[Aykut Koc Lab - Bilkent @KocLab_Bilkent]

1- In the paper “Semantic Change Detection with Gaussian Word Embeddings” published at IEEE Transactions on Audio, Speech and Language Processing, KocLab has developed the first Gaussian word embedding (w2g)-based approach to the problem of lexical semantic change detection.