Universitas Dian Nuswantoro (UDINUS), with Nurchim, Muljono, Edi Noersasongko, Ahmad Zainul Fanani, and Deshinta Arrova Dewi as activity participants, conducted the journal activity titled A Few-Shot Semantic Meta-Learning Framework with CRF for Skill Entity Recognition in Open Innovation Ecosystems during the 2026 period. The activity addressed the accelerating pace of digital transformation, changes in labour-market dynamics, the emergence of new competencies, and the need for scalable skill-intelligence systems within open innovation ecosystems.
The activity focused on Indonesian Named Entity Recognition for skill-extraction tasks in low-resource contexts. The source identifies limited research in this area, particularly where annotated data are scarce and novel skill expressions evolve rapidly. UDINUS introduced the Few-Shot Semantic Meta-Learning framework with CRF, abbreviated as FSM-CRF, for Indonesian skill entity recognition. The framework integrates semantic span representations, episodic meta-learning, and BIO-constrained CRF decoding. These components were intended to improve prototype stability and entity-boundary precision for complex, multi-token skill expressions.
Using the NERSkill.id dataset, the activity evaluated the proposed model under a 3-way, 10-shot episodic setting. FSM-CRF achieved a micro-F1 of 73.84% and outperformed the traditional supervised approaches identified in the source, namely IndoBERT fine-tuning and BiLSTM-CRF, as well as existing few-shot baselines. Ablation experiments showed that semantic span modelling and structured CRF inference played pivotal roles in improving robustness. The source also reports that meta-learning strengthened adaptability across diverse and evolving skill categories.
In relation to SDG 4, SDG 8, and SDG 9, the activity connects learning and skills research, workforce-related competency mapping, and innovation-oriented digital infrastructure. From an open innovation perspective, the framework is presented as a data-efficient solution for dynamic competency mapping, reducing dependence on costly annotation pipelines and enabling continuous updates to workforce skill taxonomies. The findings position semantic meta-learning as a promising foundation for skill-intelligence infrastructures supporting AI-enabled innovation management, strategic workforce planning, and evidence-informed policy design. This activity demonstrates UDINUS engagement with applied NLP research addressing evolving skill information in open innovation ecosystems.