arXiv subjects

Onur Gungor

Publications and source records attributed to Onur Gungor.

7 recordsLinked to original sources

arXiv2026

TurkBench: A Benchmark for Evaluating Turkish Large Language Models

With the recent surge in the development of large language models, the need for comprehensive and language-specific evaluation benchmarks has become critical. While significant progress has been made in evaluating English-language models, benchmarks for other languages, particularly those with unique linguistic characteristics such as Turkish, remain less developed. Our study introduces TurkBench, a comprehensive benchmark designed to assess the capabilities of generative large language models in the Turkish language. TurkBench involves 8,151 data samples across 21 distinct subtasks. These are organized under six main categories of evaluation: Knowledge, Language Understanding, Reasoning, Content Moderation, Turkish Grammar and Vocabulary, and Instruction Following. The diverse range of tasks and the culturally relevant data would provide researchers and developers with a valuable tool for evaluating their models and identifying areas for improvement. We further publish our benchmark for online submissions at https://huggingface.co/turkbench

cs.CL↗

arXiv2024

Improving Pinterest Search Relevance Using Large Language Models

To improve relevance scoring on Pinterest Search, we integrate Large Language Models (LLMs) into our search relevance model, leveraging carefully designed text representations to predict the relevance of Pins effectively. Our approach uses search queries alongside content representations that include captions extracted from a generative visual language model. These are further enriched with link-based text data, historically high-quality engaged queries, user-curated boards, Pin titles and Pin descriptions, creating robust models for predicting search relevance. We use a semi-supervised learning approach to efficiently scale up the amount of training data, expanding beyond the expensive human labeled data available. By utilizing multilingual LLMs, our system extends training data to include unseen languages and domains, despite initial data and annotator expertise being confined to English. Furthermore, we distill from the LLM-based model into real-time servable model architectures and features. We provide comprehensive offline experimental validation for our proposed techniques and demonstrate the gains achieved through the final deployed system at scale.

cs.IR↗

arXiv2017

Morphological Embeddings for Named Entity Recognition in Morphologically Rich Languages

In this work, we present new state-of-the-art results of 93.59,% and 79.59,% for Turkish and Czech named entity recognition based on the model of (Lample et al., 2016). We contribute by proposing several schemes for representing the morphological analysis of a word in the context of named entity recognition. We show that a concatenation of this representation with the word and character embeddings improves the performance. The effect of these representation schemes on the tagging performance is also investigated.

cs.CL↗

arXiv2017

Numerical Facet Range Partition: Evaluation Metric and Methods

Faceted navigation is a very useful component in today's search engines. It is especially useful when user has an exploratory information need or prefer certain attribute values than others. Existing work has tried to optimize faceted systems in many aspects, but little work has been done on optimizing numerical facet ranges (e.g., price ranges of product). In this paper, we introduce for the first time the research problem on numerical facet range partition and formally frame it as an optimization problem. To enable quantitative evaluation of a partition algorithm, we propose an evaluation metric to be applied to search engine logs. We further propose two range partition algorithms that computationally optimize the defined metric. Experimental results on a two-month search log from a major e-Commerce engine show that our proposed method can significantly outperform baseline.

cs.IR↗

arXiv2014

Secret Key Generation Via Localization and Mobility

We consider secret key generation from relative localization information of a pair of nodes in a mobile wireless network in the presence of a mobile eavesdropper. Our problem can be categorized under the source models of information theoretic secrecy, where the distance between the legitimate nodes acts as the observed common randomness. We characterize the theoretical limits on the achievable secret key bit rate, in terms of the observation noise variance at the legitimate nodes and the eavesdropper. This work provides a framework that combines information theoretic secrecy and wireless localization, and proves that the localization information provides a significant additional resource for secret key generation in mobile wireless networks.

cs.IT↗

arXiv2013

On the Secrecy Capacity of Block Fading Channels with a Hybrid Adversary

We consider a block fading wiretap channel, where a transmitter attempts to send messages securely to a receiver in the presence of a hybrid half-duplex adversary, which arbitrarily decides to either jam or eavesdrop the transmitter-to- receiver channel. We provide bounds to the secrecy capacity for various possibilities on receiver feedback and show special cases where the bounds are tight. We show that, without any feedback from the receiver, the secrecy capacity is zero if the transmitter-to-adversary channel stochastically dominates the effective transmitter-to-receiver channel. However, the secrecy capacity is non-zero even when the receiver is allowed to feed back only one bit at the end of each block. Our novel achievable strategy improves the rates proposed in the literature for the non-hybrid adversarial model. We also analyze the effect of multiple adversaries and delay constraints on the secrecy capacity. We show that our novel time sharing approach leads to positive secrecy rates even under strict delay constraints.

cs.IT↗

arXiv2011

Secrecy Outage Capacity of Fading Channels

This paper considers point to point secure communication over flat fading channels under an outage constraint. More specifically, we extend the definition of outage capacity to account for the secrecy constraint and obtain sharp characterizations of the corresponding fundamental limits under two different assumptions on the transmitter CSI (Channel state information). First, we find the outage secrecy capacity assuming that the transmitter has perfect knowledge of the legitimate and eavesdropper channel gains. In this scenario, the capacity achieving scheme relies on opportunistically exchanging private keys between the legitimate nodes. These keys are stored in a key buffer and later used to secure delay sensitive data using the Vernam's one time pad technique. We then extend our results to the more practical scenario where the transmitter is assumed to know only the legitimate channel gain. Here, our achievability arguments rely on privacy amplification techniques to generate secret key bits. In the two cases, we also characterize the optimal power control policies which, interestingly, turn out to be a judicious combination of channel inversion and the optimal ergodic strategy. Finally, we analyze the effect of key buffer overflow on the overall outage probability.

cs.IT↗