In the present scenario, Automatic Text Summarization (ATS) is in great demand to address the ever-growing volume of text data available online to discover relevant information faster. In this research, the ATS methodology is proposed for the Hindi language using Real Coded Genetic Algorithm (RCGA) over the health corpus, available in the Kaggle dataset. The methodology comprises five phases: preprocessing, feature extraction, processing, sentence ranking, and summary generation. Rigorous experimentation on varied feature sets is performed where distinguishing features, namely- sentence similarity and named entity features are combined with others for computing the evaluation metrics. The top 14 feature combinations are evaluated through Recall-Oriented Understudy for Gisting Evaluation (ROUGE) measure. RCGA computes appropriate feature weights through strings of features, chromosomes selection, and reproduction operators: Simulating Binary Crossover and Polynomial Mutation. To extract the highest scored sentences as the corpus summary, different compression rates are tested. In comparison with existing summarization tools, the ATS extractive method gives a summary reduction of 65%.
The computational demands of Large Language Models (LLMs) during inference are primarily determined by response length. A substantial portion of the computational cost of LLM inference arises from unnecessary autoregressive decoding, where models generate verbose explanations from simple prompts even when simple, structured answers are sufficient. We introduce a question-type-aware inference framework for prompt routing to produce compact, token-optimised responses. Our approach employs a lightweight BERT classifier to assign incoming queries to predefined question types (e.g., yes/no/maybe, multiple-choice, or descriptive). Each query is then routed to a specialised prompt template designed for concise response generation. We tested our framework on factual tasks (yes/no/maybe and multiple-choice) to ensure rigorous accuracy verification. The methodology is validated across three medical benchmark using four LLMs with 8–70 billion parameters. Results demonstrate a substantial decrease in both average token counts and floating-point operations (FLOPs), achieving peak savings of 99.4%, while maintaining baseline accuracy. This approach offers a scalable solution to lower the operational costs of deploying LLMs in specialized, knowledge-intensive domains.