Knowledge Extraction and Quality Inspection of Chinese Petrographic Description Texts with Complex Entities and Relations Using Machine Reading and Knowledge Graph: A Preliminary Research Study

Round 1

Reviewer 1 Report

Dear authors,

in my opinion this manuscript dealing with the knowledge detection from petrographic description texts with complex entity and relation is very interesting, well organized and the results are  clearly presented. Moreover, I suggest for publication for this journal and I'm interested to see your  future investigation on the relationship extraction method . 

Kind regards

Author Response

Thank you very much for your comments and suggestions. The manuscript which focus on complex entity and relation extraction method is writing and we hope to get more suggestions from you in the future.

Reviewer 2 Report

Review minerals-1772137 Chen et al.

Knowledge Detection from Petrographic Description Texts with complex Entity and Relation using Machine Reading and Knowledge Graph

by Zhongliang Chen, Feng Yuan, Xiaohui Li, Xiang Wang, He Li, Bangcai Wu and Yuheng Chen

 

In general the English is not too good (although I am not native speaker I would say so), it needs some editing, I picked up a lot of spelling mistakes, and sentences are to long or with lots of routine repetitions.

In detail:

Detailed comments:

Title: "c" in "complex entity" uses a different fond, should be same...

line 11: (1) Background: "Geological survey...", should be "(1) Background: The Geological Survey of the Anhui Province..."

line 13: "in many domains", must be plural

line 15: "were defined based on the petrological...", "the" missingLine 33: "...domains..." (plural)

line 34: "respond"

line 37: "helping to make decisions by means of deep information..."

line 40: instead of "more and more" "increasingly"

line 41+42: not a complete sentence: "PaleoDeepDive is a ...database [15]"

line 78: "In General professionally trained geologists..."

line 82: "...described...Chinese...", incomplete sentence: Metamorphic

line 89+90: "... has just been designed...", Is "moudle" the right word? rather "coupling"

line 114: "...graph..."

line 161: "...petrographic..."

line 182: "...approachers.."

line 188: "The models, such as DBN have been carrying out experiments..." carrying

line 191: "...according to specific tasks."

line 192: "some approaches.."

line 193/194: "...have begun to identify the geologically named entities in the domain of geosciences."

line 195: "...having more choices"

line 197: between the bi-directional..."

line 200: "...petrographic description."

line 201: "...a format file..." "a"?

line 209: Framework for the NER models comparison. What means NER?

line 211: "As mentioned above..."

line 272: "...detail in another article."

line 308: "accessory mineral"

line 350: "triples match,..." (plural)

line 357: "...If the numble is only one, the process comes to an end."

line 373/375: Table 2/Table 3: Header partly in Chinese, I do not understand, wonder whether it should be all in English...

line 377: Figure 8b and tables 2+3, aren't there minerals missing such as kyanite, olivine as minor mineral (MI_MINERAL), accessory mineral (ACC_MINERAL), and the groundmass category "groundmass mineral (GRO_MINERAL)" are the main minerals of the eclogite not phenocrystic minerals (PHE_MINERAL)?

line 392: "...F1 scores", "...particular, the F1 value of..."

line 399: F1 values of the BERT–BiLSTM–CRF model against the BERT–CRF model and BiLSTM–CRF model are not that different in performance score. Does it make a great difference?

line 412: "...entities..."

line 413: "...knowledge graph." (gap)

line 425: "application"

line 434: "framework"

line 446 and 447: I strongly contradict the conlusion 3) "Based on knowledge graph, knowledge detection and recommendation can improve the quality of rock description."

The method relies on the quality of the prior knowledge of the metamorphic plutonic rocks in the Fengxiangyi sheet; because the primary petrographic description is the most essential step on rock characterizazion, if this is wrong or faulty, all following reasoning is also faulty. The danger is that the automated alorithms will also interpret faulty primary information because it does not perform an optical comparison e.g. with a scan of the thin section of the specific rock.

Criticism:

I cannot comment on the workflow of the framework for the complex petrographical knowledge detection, and trust that the petrographic relation extraction based on enriching R-Transformer model and the knowledge extraction and consistency calculation is done in a plausible way (figure 7)

To me it is not clear if figure 6 shows the geological map with prior petrogrpahic information of after knowledge detection from petrographic description texts with complex entity and relation using machine reading and knowledge graph? If it does, then for me it would be more useful to test of the method is able to distinguish the various gneiss types (which do not differ a lot but which are difficult to distinguish) than to differentiate eclogite from granitic gneiss.

However I disagree that this study takes the extraction and detection of complex petrographic knowledge a step further, since e.g. the petrography of eclogite is generalised too much so it can not be distinguised from a metabasite near earth surface.

Overall an interesting contribution to knowledge Detection from Petrographic Description Texts with complex Entity and Relation using Machine Reading and Knowledge Graph, however especially textures and structures in rocks are difficult to schematise and recognize automatically, The quality of the knowledge detection can only be as good as the primary petrographic description texts, so regular spot checks or random samples would be necessary. I miss a bit of methodology criticism.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Dear Authors, the process of automatic data mining based on deep learning is an interesting issue, very relevant nowadays. I found this text interesting and characterizing the preliminary research. 

I know that in the text of the article was mentioned indexing the names and assigning them to the Chinese equivalents, but in the table, in addition to the low chi, perhaps it would be worthwhile to enter the English equivalents under these names in parentheses? I think this would be much more convenient for the reader.

It is worth mentioning that searching for texts in google is related to their positioning and does not always give accurate results. Maybe it would be worth referring to other databases as well? It is also worth mentioning how authors adapted the search material, whether they uploaded articles to the software themselves or let it automatically search the Internet, because this can also have an impact on the quality and quantity of results. On the other hand, some texts are not available and other descriptions may be misleading. It might be worth writing something more about this. 

In the case of semantic definition of various phrases, I admit that the structure, texture is not a very good illustration of the search, because these are words that have different meanings in different languages and are not always accurately used by geologists themselves. Will it then be possible to discover from the context the fact of the reversed use of these terms?

In addition, I would like to point out that the literature items are also relatively few, I propose to add also those items that are related to deep learning and the process of searching and creating databases.

There are a lot of these issues and I think I would be inclined to add in the title of the text that this is preliminary research.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 4 Report

- Send the text to a native English-speaker;

- The Abstarct needs extensive modification to clarify the ideas. It is confused and senseless;

-The Table 2 has chinese characters;

-The Conclusions do not grasp the entire text. Try to modify accordingly;

Author Response

Point 1: Send the text to a native English-speaker;

Response 1: This revised manuscript has been send to MogoEdit (https://www.mogoedit.com) for its English editing provided by native English-speaker.

Point 2: The Abstarct needs extensive modification to clarify the ideas. It is confused and senseless; The Conclusions do not grasp the entire text. Try to modify accordingly;

Response 2: The Abstract and Conclutions have been revised according to the comments.

Point 3: The Table 2 has chinese characters;

Response 3: In Table 1–3, the terms in Chinese are the English equivalents and have been entered in parentheses. This modification would be much more convenient for the reader.

Round 2

Reviewer 2 Report

In general the language could still be improved by a physical native speaker/editor if at hand.

The following orthographic/grammatical changes are suggested:

line 12/13: Geological surveying is undergoing a digital transformation process towards the adoption of intelligent methods in China

line 118/119: Suggestion to rephrase: At present, digital geological survey has published for China, and an kognitive geological survey is also under development.

line 390: In an area..., ...,e.g. at 1:200 000 scale ...

line 406: The characteristics of rock composition are described by means of major and minor minerals, and accessory minerals, respectively.

line 507: mineral resources

line 510: separated

there may be more, as said previously it would be good if a native speaker would go through it and edit again.

 

 

Author Response

Point 1: In general the language could still be improved by a physical native speaker/editor if at hand. there may be more, as said previously it would be good if a native speaker would go through it and edit again.

Response 1: This revised manuscript has been send to MogoEdit (https://www.mogoedit.com) for its native English editing once more. Please see the certification of English editing in the attachment.

 

Point 2: The following orthographic/grammatical changes are suggested:

line 12/13: Geological surveying is undergoing a digital transformation process towards the adoption of intelligent methods in China

line 118/119: Suggestion to rephrase: At present, digital geological survey has published for China, and an kognitive geological survey is also under development.

line 390: In an area..., ...,e.g. at 1:200 000 scale ...

line 406: The characteristics of rock composition are described by means of major and minor minerals, and accessory minerals, respectively.

line 507: mineral resources

line 510: separated

Response 2: The manuscript has been revised according to this suggestions.

Author Response File: Author Response.pdf

Reviewer 4 Report

After several improvements, the manuscript has a good shape.

Author Response

Point: After several improvements, the manuscript has a good shape.

Response: Thank you again for your comments and suggestions.