Research on Intelligent Question-And-Answer Multi-Intention Analysis Method of Converter Steelmaking Knowledge Graph

Addressing the issues of information dispersion, reliance on human experience, and low data utilization efficiency in the field of converter steelmaking, a knowledge graph intelligent question answering multi-intentanalysis method based on the roberta-bigru-localattention model was proposed. Firstly, robustly optimized BERT approach was used to pre-trained language model to extract multi-level semantic features in feature extraction module, and dynamic weighted fusion was introduced to enhance domain term representation ability. Secondly, the bi-directional gated recurrent unit was used to capture the context dependence in the encoding module, and the fusion vector output by feature extraction was converted into a hidden state sequence containing bidirectional semantic dependence; Finally, the decoding module combines the local attention mechanism to dynamically focus on key information, strengthen the association constraints between adjacent operation parameter labels, solve the semantic coupling problem of multi-label classification, and break through the limitations of the traditional single-intention parsing framework. Compared with LSTM, BiGRU simplifies the gating structure, reducing computational complexity while still being able to effectively capture long-term dependencies. In the converter steelmaking scenario, where the text length is moderate, BiGRU achieves a better balance between performance and efficiency. Experimental results show that the RBL-BiLA (Roberta-bigru-local attention) model achieves a precision rate of 94.12%, a recall rate of 95.33%. And an F1 score 95.63% significantly outperforms the comparative model, and ablation experiments verify the effectiveness of dynamic fusion and local attention mechanisms. This research provides technical support for intelligent knowledge management in converter steelmaking processes, enhances the predictive ability of steel type attribute association mining and new steel type development, and offers a scalable solution for multi-intent parsing tasks in industrial scenarios.