Politically Unbiased News Report Generation

Political bias in digital news has raised significant concerns about media polarization and its impact on public discourse. This paper introduces a novel politically unbiased news report generation framework that leverages the power of large language models (LLMs) and transformers for creating balanced news content. Recognizing the alarming decline of journalistic integrity in India, our work specifically addresses the need for neutral news dissemination in the Indian context. The proposed system integrates a multi-step approach, starting with the classification of news articles as left-leaning or rightleaning, followed by the detection of biased terminology using fine-tuned models. Identified biases are then mitigated through adaptive filtering techniques to ensure impartiality. These processed inputs are subsequently fed into a LoRA fine-tuned LLaMA model trained on a corpus of neutral news articles to generate unbiased reports. Experimental results demonstrate the effectiveness of the framework in significantly reducing bias in the generated content compared to traditional automated text generation methods. This work contributes to advancing neutral news dissemination and offers a promising tool for countering media polarization in politically sensitive environments.