Multi-Feature Fusion Method for Chinese Pesticide Named Entity Recognition

Round 1

Reviewer 1 Report

Summary of the paper:

The authors proposed BiLSTM-IDCNN-CRF model for Chinese pesticide named entity recognition.

Strong point:

1.     The motivation of this paper is good.

2.     The paper is well organized and easy to read.

3.     Experiments conducted on a reasonable sized dataset which showed the effectiveness of the proposed method.

Weak points:

1.     The limitations of existing works are not explicitly clear.

2.     Please mention the tools you used for preprocessing with proper references.

3.     I like to see a paragraph mentioning how the proposed model deals with long texts.

4. The quality of the figures needs to be improved.

Author Response

Dear Reviewer，

On behalf of my co-authors, we are very grateful to you for giving us an opportunity to revise our manuscript.

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper presents a method to tackle the NER task from Chinese pesticide information. The method is based on a BiLSTM-IDCNN-CRF model. The components of the method and the intuitive ideas about their use are well explained. The paper shows an extensive background and related work. However, no comparison is made with the proposed method. New contributions or ideas should be highlighted and explained.

The paper presents several areas for improvement:

• I suggest highlighting the relevance of the pesticide NER task.

• The abstract contains acronyms that are not well known such as IDCNN, which is expanded until the related work.

• The main contributions in the introduction appear to be descriptions of components rather than contributions.

• According to the authors, the architecture BiLSTM-IDCNN-CRF has been used in the work in reference [23] in ecological NER, what is the difference with the proposed method? is it an adaptation for the pesticide NER task? Hence, the contribution should be highlighted

• Could the results be compared with a similar model BiLSTM-CNN-CRF, which has been used in the NER task?

• I suggest adding the English translation of the data column in Table 1

• I consider that more details of the dataset are necessary. For example, the number of instances by category. This is important because an error analysis by category was reported.

• Tables 2 y 4 have 10 categories, but the text mentioned 11. Which is the missing category?

Author Response

Dear Reviewer，

On behalf of my co-authors, we are very grateful to you for giving us an opportunity to revise our manuscript.

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The questions have been answered.  Some experiments and paragraphs have been added to improve the paper's content.