A Named Entity and Relationship Extraction Method from Trouble-Shooting Documents in Korean
Round 1
Reviewer 1 Report
Authors proposed a named entity recognition (NER) approach based on dependency parsing for equipment maintenance documents. Although there are relevant works in literature that used NER and dependency parsing, authors did not clearly specify the novelty of the proposed work. Therefore, in my opinion, the novelty of the work is low. Authors should explain their contribution in clear sentences both in the abstract and introduction.
Other than this, technical details are not explained in detail such as conditional random field, Bidirectional encoder representation from transforms, which the authors used in their works. In addition, in many cases acronyms are used then their explanations are given in the following pages such as KLUE-BERT and KLUE-DP. This makes the reading harder.
Implementation details are not given. Which software did you use?
Evaluation setup and metrics are not explained, which makes it difficult to follow the results chapter. How did you compute precision and recall in the context of tokens, please provide the formulas.
Overall, the papers need throughout editing to improve its technical soundness and explanations.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 2 Report
The paper presents an extraction method from equip-2 ment maintenance documents in Korean, based on the use of Named Entities Recognition(NER) and rule-based approaches.
The paper introduces in sections 1 and 2 the problem and the related work. From my point of view, the problem is clearly stated whereas the related work is not so clear. In particular, it introduces previous work in the field but does not state what those approaches are not adequate for the field the paper is focused on or what are the advantages or main contributions of the proposal made in the paper.
Sections tree to six introduce the system proposed, explaining all the steps done for the extraction. Again, it is not clear what are the contributions or novelties of the paper. The description provided details the process and techniques used, but do not indicate what are the advantages or improvements that the method achieves.
Section 7 introduces the experiments and analyzes the results. It is not clear, whether a known corpus of documents is used or not. It is also unclear how the evaluation is done. The documents were previously classified? There has been experts who do the evaluation? In particular, the Tables 3 and 4, introduce results whose origin is not clear.
Conclusion sections are short and may be improved by adding more detail in what are the contributions of the paper and the proposal. Moreover, it is important to introduce what are the drawbacks of the system and the future work.
Summarizing, the paper may be of interest, but it must be improved to make clear what are the main contributions and how the evaluation has been made.
Finally, a general revision of the English language must be done. Although, the paper is easy to follow and understandable, there are some grammatical errors that can be improved.
Author Response
Please see the attachment
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
Authors substantially improved the manuscript after major revisions. My main concern was the novelty of the proposed work and technical details was missing. Although the proposed named entity recognition using NER and dependency parsing is not that much novel, its application areas seems novel. In addition, authors added several details about technical details in the revised manuscript.