Dataset for Knowledge Retrieval in Digital Enterprises Using BERT–GNN AI Workflow

This dataset supports research on AI-driven knowledge retrieval in digital enterprises and was developed for evaluating a hybrid artificial intelligence workflow combining Bidirectional Encoder Representations from Transformers (BERT) and Graph Neural Networks (GNNs). The dataset integrates both structured and unstructured enterprise data collected from multiple sectors, including information technology, manufacturing, healthcare, and education. The structured component consists of survey responses from more than 250 professionals actively involved in digital transformation, knowledge management (KM), artificial intelligence, and big data analytics initiatives. The questionnaire includes 24 Likert-scale items (1–5 scale) covering organizational agility, operational efficiency, AI adoption in KM, big data analytics impact on knowledge sharing, and user-perceived system effectiveness. The unstructured component includes semi-structured interview transcripts from 30 domain experts, totaling over 120,000 words, along with supplementary enterprise materials such as company annual reports, digital transformation strategy documents, and 15 industry case studies. Additional anonymized internal knowledge-base documents were incorporated to enrich the representation of organizational knowledge flows. All records were collected during a defined study period and anonymized in accordance with institutional ethical guidelines to ensure confidentiality and compliance. The dataset enables training and evaluation of deep learning models for enterprise knowledge retrieval, semantic knowledge representation, and digital transformation analytics, particularly frameworks integrating transformer-based language models and graph neural network architectures.