Gjergji Kasneci


2024

pdf bib
Explainability Meets Text Summarization: A Survey
Mahdi Dhaini | Ege Erdogan | Smarth Bakshi | Gjergji Kasneci
Proceedings of the 17th International Natural Language Generation Conference

Summarizing long pieces of text is a principal task in natural language processing with Machine Learning-based text generation models such as Large Language Models (LLM) being particularly suited to it. Yet these models are often used as black-boxes, making them hard to interpret and debug. This has led to calls by practitioners and regulatory bodies to improve the explainability of such models as they find ever more practical use. In this survey, we present a dual-perspective review of the intersection between explainability and summarization by reviewing the current state of explainable text summarization and also highlighting how summarization techniques are effectively employed to improve explanations.

pdf bib
P-TA: Using Proximal Policy Optimization to Enhance Tabular Data Augmentation via Large Language Models
Shuo Yang | Chenchen Yuan | Yao Rong | Felix Steinbauer | Gjergji Kasneci
Findings of the Association for Computational Linguistics ACL 2024

A multitude of industries depend on accurate and reasonable tabular data augmentation for their business processes. Contemporary methodologies in generating tabular data revolve around utilizing Generative Adversarial Networks (GAN) or fine-tuning Large Language Models (LLM). However, GAN-based approaches are documented to produce samples with common-sense errors attributed to the absence of external knowledge. On the other hand, LLM-based methods exhibit a limited capacity to capture the disparities between synthesized and actual data distribution due to the absence of feedback from a discriminator during training. Furthermore, the decoding of LLM-based generation introduces gradient breakpoints, impeding the backpropagation of loss from a discriminator, thereby complicating the integration of these two approaches. To solve this challenge, we propose using proximal policy optimization (PPO) to apply GANs, guiding LLMs to enhance the probability distribution of tabular features. This approach enables the utilization of LLMs as generators for GANs in synthesizing tabular data. Our experiments demonstrate that PPO leads to an approximately 4% improvement in the accuracy of models trained on synthetically generated data over state-of-the-art across three real-world datasets.

pdf bib
Is Crowdsourcing Breaking Your Bank? Cost-Effective Fine-Tuning of Pre-trained Language Models with Proximal Policy Optimization
Shuo Yang | Gjergji Kasneci
Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024)

Wide usage of ChatGPT has highlighted the potential of reinforcement learning from human feedback. However, its training pipeline relies on manual ranking, a resource-intensive process. To reduce labor costs, we propose a self-supervised text ranking approach for applying Proximal-Policy-Optimization to fine-tune language models while eliminating the need for human annotators. Our method begins with probabilistic sampling to encourage a language model to generate diverse responses for each input. We then employ TextRank and ISODATA algorithms to rank and cluster these responses based on their semantics. Subsequently, we construct a reward model to learn the rank and optimize our generative policy. Our experimental results, conducted using two language models on three tasks, demonstrate that the models trained by our method considerably outperform baselines regarding BLEU, GLEU, and METEOR scores. Furthermore, our manual evaluation shows that our ranking results exhibit a remarkably high consistency with that of humans. This research significantly reduces training costs of proximal policy-guided models and demonstrates the potential for self-correction of language models.

2014

pdf bib
BEL: Bagging for Entity Linking
Zhe Zuo | Gjergji Kasneci | Toni Gruetze | Felix Naumann
Proceedings of COLING 2014, the 25th International Conference on Computational Linguistics: Technical Papers