Corpus Design and Extractive Speech Summarization for Low- Resource Languages

Corpus Design and Extractive Speech Summarization for Low-Resource Languages delivers a comprehensive analysis of modern natural language processing and speech technology frameworks tailored for data-constrained linguistic environments. As automatic speech recognition and text summarization systems advance, building reliable computational models for languages lacking extensive annotated corpora remains a critical challenge. This monograph establishes the theoretical foundations, architectural strategies, and data engineering methodologies required to construct effective speech corpora and implement robust extractive summarization pipelines.

The text systematically covers acoustic and textual data collection protocols, phonetic alignment, data augmentation strategies, and noise reduction techniques for low-resource speech processing. It explores graph-based algorithms, neural sentence scoring, acoustic feature extraction, and salient phrase extraction designed to identify key informational units directly from spoken audio and under-annotated transcriptions. Designed for computational linguists, speech processing engineers, and natural language processing researchers, this volume provides the analytical tools needed to scale language technologies to under-represented global languages.