EIED: A Curated Dataset for Knowledge Extraction in Electronic Information Equipment Domain

As a core task in web data mining, knowledge extraction holds significant application value in knowledge graph construction. Addressing the lack of domain-specific features in existing public datasets for electronic information equipment applications, this study proposes a domain-specific dataset construction methodology based on a collaborative framework of “multi-source collection, heterogeneous processing, and semantic annotation.” We successfully developed EIED, a specialized dataset for electronic information equipment. This dataset integrates multi-modal heterogeneous data sources, including technical literature, documentation, equipment parameters, and industry reports, forming a structured corpus of $200,000-$ character scale after semantic cleansing and expert annotation. Experimental validation employed a bidirectional comparative mechanism: BERT, BERT-LSTM, and BERT-BiLSTM hybrid models were tested on both EIED and generic datasets for named entity recognition and relation extraction tasks. Results demonstrate that models trained with EIED significantly outperform those using generic datasets in both entity recognition and relation extraction. This research establishes a high-quality data foundation for subsequent knowledge extraction and graph-based development in electronic information equipment applications.