JEEMRC: Joint Event Detection and Extraction via an End-to-End Machine Reading Comprehension Model

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

In this manuscript, the authors create a new Joint Event Extraction via end-to-end Machine Reading Comprehension (JEEMRC) model for event extraction from a given sentence. The advantage of the proposed model, compared to existing approaches, is that it jointly solves the two subtasks (event detection and argument extraction) to alleviate error propagation. In this way, it utilizes interaction information between event types and argument roles. Experimental results confirm the effectiveness of this new event extraction technique.

The topic is interesting and worth investigating. The manuscript is well organized and well written.

 

My remarks are as follows:

For better reproducibility of the proposed method, please provide a link to your programming code.

In the ‘4. Analysis’ section, time complexity of authors’ model should be compared with complexity of existing analogs for event extraction.

In ‘5. Conclusion’ section, study limitations should be commented.

 

Technical remark:

Figure 4:

‘f1_e’ -> ‘F1_EC’,

‘f1_i’ -> ‘F1_AI’,

‘f1_c’ -> ‘F1_AC’.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This paper develops a joint event detection and extraction scheme through an end-to-end machine reading comprehension model. The following comments should be addressed:

- Abstract, line 3, this sentence seems to conflict with the previous "the success of MRC". It is suggested to highlight how ignoring the correlation between event types and argument roles can influence the outcome. Since this is the motivation.

- Introduction, the first paragraph should highlight the background and motivation. Where can this work be applied? What gaps are there? For example, the problem of "ignoring correlation ..." is not stated here.

- Figure 1, it is suggested to include an illustration of how traditional methods would do to the sentence, with key differences highlighted. This helps the reader to understand the novelty of the author's work.

- In section 1.3, the title, "IE" is not clear to the readers.

- In section 2.3.1, please elaborate more on the model architecture, including BERT. 

- In Section 3, the table titles are too general: Table 4, "Experiment results", what results are they? Similar issues can be seen in Tables 5, 6, and 9. It is hard to understand the contents solely based on the tables.

Comments on the Quality of English Language

- Abstract: Line 5, this sentence is not quite clear, please revise.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

All comments have been resolved.