TR-EduVSum: A Turkish-Focused Dataset and Consensus Framework for Educational Video Summarization

TR-EduVSum sits at the intersection of low-resource NLP (Turkish) and specialized domain tasks (educational summarization). Its primary value is not the code itself, but the dataset: 3,281 human-authored summaries for 82 videos represents a significant manual effort that provides 'data gravity.' The AutoMUP framework attempts to solve the 'consensus' problem in summarization—how to create a single ground truth from noisy human inputs. From a competitive standpoint, its defensibility (4) is moderate; while the dataset is a moat against generic models, the sample size (82 videos) is relatively small for industrial applications. Frontier labs like OpenAI or Google are unlikely to target this niche specifically (low frontier risk), as they prioritize general-purpose multi-lingual performance. However, local Turkish AI initiatives or specialized EdTech platforms (like Khan Academy's Turkish branch or local LMS providers) could easily replicate or absorb this logic. The lack of stars (0) and very recent age (9 days) indicate this is currently an academic output with no commercial momentum yet. The 'displacement horizon' of 1-2 years reflects the rapid improvement of zero-shot Turkish capabilities in LLMs, which may eventually render specialized consensus frameworks less critical than they are today.