A Semantic Annotation Pipeline towards the Generation of Knowledge Graphs in Tribology

Round 1

Reviewer 1 Report

In this manuscript, in order to overcome the ‘knowledge acquisition bottleneck’ problem, authors drew on the idea of Gene Ontology database in biology, and tried to use natural language processing (NLP) techniques to annotate publications in tribology and generate knowledge graphs. Authors built a BERT-based pipeline and it can extract tribological information from publications in PDF format automatically or semi-automatically. Although the work was significant and the preliminary results confirmed the feasibility, I have the following questions and suggestions:

• More technical details should be provided (maybe in supplementary materials). For example, in the document analysis module, authors trained three pre-training models BERT, SciBERT and SpanBERT on tribological data and then compared their performance. But details of the training was missing, such as what method was used, and what was the computational cost?
• I also tried to test the pipeline and downloaded the code from github. It seems that the UI and the other three modules work independently of each other in different conda environments. In my test, the UI part worked fine, but I had some trouble with the other three modules. Therefore, I suggest that authors may pay some attention to usability issues, such as whether four modules can run in the same conda environment? Or build a web server for everyone to try?

Comments for author File: Comments.docx

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

the abstract must be more specific and comprehensive about the outcome of research, 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Title: A Semantic Annotation Pipeline towards the Generation of Knowledge Graphs in Tribology

 

Authors: Patricia Kügler, Max Marian, Rene Dorsch, Benjamin Schleich and Sandro Wartzack

 

This paper used semantic annotation and natural language processing techniques for the purpose of simplifying the manual research of scientific literature. The subject and the findings of the paper are interesting. However, major revisions need to be addressed before it can be considered for publication.

In the Abstract, the authors should highlight the main findings of the paper.

The motivations of the paper are not well highlighted. In other words, the authors should add more details in the literature review, such as pros and cons of each existing method, and from there deduce the necessity of this work.

The authors should verify the algorithm with more types of data, for example documents with a lot of tables.

The authors should also discuss the computational time of this algorithm. I think it is a very important factor, because when dealing with scientific literature, we are dealing with an enormous amount of input data.

The conclusion needs to be revised; this is a very important section. The main findings of the work need to be recalled in more details in this section. In addition, it is missing some perspectives. What can you do to further improve the algorithm?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

Dear authors,

I have read your manuscript with interesting. I am glad to have opportunity to review that paper. 
I suggest to consider my a few comments which could be helpful to improve your paper I hope.

Abstract

Line 22, I suggest to extend acronym BERT during first using it.

Keywords
I suggest to remove keywords like: tribology; semantic annotation; knowledge graphs. There is no justification to repeat words from the title

Introduction (shouldn't be numbered chapter - by the way)

Line 45 [4-6] three citation in one bracket seem's to be too loosely related to the issue at hand this time.

Line 67 the noun "dilemma" is used incorrectly, in fact it means an alternative, a choice between two ways

Generally I consider your paper as very valuable and almost faultless.

Regards

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

The authors have addressed the comments