MenuNER: Domain-Adapted BERT Based NER Approach for a Domain with Limited Dataset and Its Application to Food Menu Domain

Round 1

Reviewer 1 Report

This research applied known NLP algorithms to one particular use case, Food Menu Domain.
The authors have clearly presented the methods they used. 
Therefore I think that the paper is worth to publish after some additional explanations.
First of all I would like to see publiclly available annotated dataset for menu dishes. 
That is good either for result verification as well as for further research (open data concept).

4.1.3. Experiment Setting 
It will be interesting to obtain the information about training time.

4.2. Experimental Results
Table 1. - please compare executon time of various Embeddings. 
I think that the difference between obtained results (F1 score) are practically negligible.
Please comment the cases where the improvements have been achieved and why.

In any case, it would be good in more details to comment and explain the obtained results.
This is especially true of the 4.2.2. Evaluation for MenuNER.
The results are interesting but it should be explained why some model better interprets certain input data and why others miss them.

3. Conclusion and Future work (it is 5. Conclusion and Future work)

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

In general, interesting paper, easy to follow (well-written) -- though some English grammar could be improved, but it was not prohibitive to understanding the paper. And good methodology.

Author Response

Hi,

We have modified the paper to provide improved English language and phrasing wherever possible.

Thanks.

Round 2

Reviewer 1 Report

The authors responded to all my remarks and accordingly corrected the paper.