Pretrained Language–Knowledge Graph Model Benefits Both Knowledge Graph Completion and Industrial Tasks: Taking the Blast Furnace Ironmaking Process as an Example

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper proposes a two-fold approach to advance knowledge graph completion and address industrial tasks simultaneously. The methodology comprises firstly the development of a pretrained language-knowledge graph model along with dynamic task-related subgraphs, and secondly the derivation of a fine-tuning dynamic model specifically for industrial applications. The paper is well-crafted, presenting an engaging topic that aligns perfectly with the journal's focus and scope. The methodology is robust and supported by relevant literature, making the study a substantial contribution to the field.

 

However, before recommending this manuscript for acceptance, this reviewer would like to propose several minor revisions to enhance clarity and overall impact:

 

- Abstract: Include specific experimental results highlighting the study's findings and contributions.

- Figure 2: Eliminate the redundant "Decision support" element for clarity.

- Table 1: Consider removing if it does not add significant value to the paper.

- Table 2: Ensure references accompany all methods mentioned.

- Table 3 and the relavant paragraph: Provide more details on the evaluation process for graph embedding methods.

Comments on the Quality of English Language

Minor editing of English language required

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This paper addresses challenges in constructing and updating knowledge graphs for dynamic production data. It proposes a multi-task learning framework, including semantic adjacency matrix learning and entity link prediction. Expert networks and causal relationships enhance the model, while dynamic graph embedding captures the dynamic nature of blast furnace production. Simulation results validate the framework's effectiveness in introducing knowledge into the industrial context.

The paper's subject matter is captivating, and its overall structure is deemed satisfactory. Nevertheless, there are notable major concerns that need clarification.

1.  Does the proposed interactive reasoning framework for semantic mining and entity link prediction, based on multi-task learning, offer a novel and innovative solution to the challenges associated with knowledge graph construction in dynamic industrial settings?

2.  Are the two defined interrelated subtasks, namely semantic adjacency matrix learning at the ontology layer and entity link prediction at the instance layer, effectively presented as a structured and methodological foundation for the proposed framework?

3.  How impactful are the use of expert networks and causal relationships in constructing a multi-task learning framework?

4.  To what extent does the incorporation of dynamic graph embedding technology address the dynamic characteristics of the blast furnace production process, and how practical is this consideration for real-world industrial applications?

5.  Can the satisfactory results obtained in simulation experiments be quantitatively assessed or compared with existing methods, and how do these results contribute to the validation of the proposed framework's effectiveness?

6.  Do the identified contributions of the paper, such as summarizing multi-source data processing methods and introducing a Language-Knowledge graph model, align with the needs of the industrial context?

How does the suggested two-stage model construction strategy, involving pre-training and fine-tuning, enhance the overall understanding and applicability of knowledge graph models for industrial tasks?

7.  In terms of transparency, how might the paper benefit from a more detailed discussion of potential limitations or challenges encountered during the implementation of the proposed framework?

8.  Would providing specific quantitative metrics or further comparisons with existing methods strengthen the experimental validation of the proposed framework?

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The paper can be published in its current form.