Hamidreza Baradaran Kashani

Mehrdad Ghassabi, Spehr Rajabi, Hamidreza Baradaran Kashani et al.

Contemporary neural machine translation (NMT) systems are almost exclusively built by training on supervised parallel data. Despite the tremendous progress achieved, these systems still exhibit persistent translation errors. This paper proposes that a post-training paradigm based on reinforcement learning (RL) can effectively rectify such mistakes. We introduce a novel framework that requires only a general text corpus and an expert translator which can be either human or an AI system to provide iterative feedback. In our experiments, we focus specifically on English-to-German translation as a representative high-resource language pair. Crucially, we implement this RL-based post-training using Direct Preference Optimization (DPO). Applying our DPO-driven framework to the gemma3-1b model yields a significant improvement in translation quality, elevating it's COMET score from 0.703 to 0.747 on the English to German task. The results demonstrate that DPO offers an efficient and stable pathway for enhancing pre-trained NMT models through preference-based post-training.

Mehrdad Ghassabi, Pedram Rostami, Hamidreza Baradaran Kashani et al.

The rapid advancement of language models has demonstrated the potential of artificial intelligence in the healthcare industry. However, small language models struggle with specialized domains in low-resource languages like Persian. While numerous medical-domain websites exist in Persian, no curated dataset or corpus has been available making ours the first of its kind. This study introduces a newly curated dataset comprising 20k doctor-patient Q\&A pairs and 60\% of a 90-million-token crawled corpus from medical magazines. Using a parameter-efficient fine-tuning approach, we enhanced the medical knowledge of the baseline model, aya-expanse-8b. Benchmark evaluations demonstrate that the fine-tuned model achieves improved accuracy in medical question answering and successfully passed the Iranian Basic Medical Science Entrance Exam (IBSEE) in September 2023, which the baseline model did not. Additionally, the fine-tuned model improved Persian-translated MMLU accuracy by an average of 2.67\%. This work highlights the potential of leveraging open-access online data to enrich small language models in medical fields, providing a novel solution for Persian medical AI applications suitable for resource-constrained environments. Future research could explore multimodal input to further enhance performance.

Mehrdad Ghassabi, Sadra Hakim, Hamidreza Baradaran Kashani et al.

Enhancing reasoning capabilities in small language models is critical for specialized applications such as medical question answering, particularly in underrepresented languages like Persian. In this study, we employ Reinforcement Learning with AI Feedback (RLAIF) and Direct preference optimization (DPO) to improve the reasoning skills of a general-purpose Persian language model. To achieve this, we translated a multiple-choice medical question-answering dataset into Persian and used RLAIF to generate rejected-preferred answer pairs, which are essential for DPO training. By prompting both teacher and student models to produce Chain-of-Thought (CoT) reasoning responses, we compiled a dataset containing correct and incorrect reasoning trajectories. This dataset, comprising 2 million tokens in preferred answers and 2.5 million tokens in rejected ones, was used to train a baseline model, significantly enhancing its medical reasoning capabilities in Persian. Remarkably, the resulting model outperformed its predecessor, gaokerena-V, which was trained on approximately 57 million tokens, despite leveraging a much smaller dataset. These results highlight the efficiency and effectiveness of reasoning-focused training approaches in developing domain-specific language models with limited data availability.

Hamidreza Baradaran Kashani, Ata Jodeiri, Mohammad Mohsen Goodarzi et al.

Quality and intelligibility of speech signals are degraded under additive background noise which is a critical problem for hearing aid and cochlear implant users. Motivated to address this problem, we propose a novel speech enhancement approach using a deep spectrum image translation network. To this end, we suggest a new architecture, called VGG19-UNet, where a deep fully convolutional network known as VGG19 is embedded at the encoder part of an image-to-image translation network, i.e. U-Net. Moreover, we propose a perceptually-modified version of the spectrum image that is represented in Mel frequency and power-law non-linearity amplitude domains, representing good approximations of human auditory perception model. By conducting experiments on a real challenge in speech enhancement, i.e. unseen noise environments, we show that the proposed approach outperforms other enhancement methods in terms of both quality and intelligibility measures, represented by PESQ and ESTOI, respectively.

Hamidreza Baradaran Kashani, Ata Jodeiri, Mohammad Mohsen Goodarzi et al.

Clipping, as a current nonlinear distortion, often occurs due to the limited dynamic range of audio recorders. It degrades the speech quality and intelligibility and adversely affects the performances of speech and speaker recognitions. In this paper, we focus on enhancement of clipped speech by using a fully convolutional neural network as U-Net. Motivated by the idea of image-to-image translation, we propose a declipping approach, namely U-Net declipper in which the magnitude spectrum images of clipped signals are translated to the corresponding images of clean ones. The experimental results show that the proposed approach outperforms other declipping methods in terms of both quality and intelligibility measures, especially in severe clipping cases. Moreover, the superior performance of the U-Net declipper over the well-known declipping methods is verified in additive Gaussian noise conditions.