A Review of AI-Driven Conversational Chatbots Implementation Methodologies and Challenges (1999–2022)

Round 1

Reviewer 1 Report

This is a very well-organized and concise review paper. Clearly this is a good asset and will get a lot of attention for those who are interested to do research on adopting Chatbots to support learning in the future. To increase the impact of this paper, I strongly suggest the authors to propose 10 research questions, instead of only 3. It may even be more impactful if the authors develop a research agenda under this topic. In sum, I strongly recommend this well-organized and timely paper to be published.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

1. the introduction needs to include more literature citations

2. if ethical issues are not the subject of this study, I do not think they need to be specifically mentioned

3. are the results in 4.1. necessary? It is not relevant to the research question

4. this paper is quite specific in its description of the analysis based on the research question. Only the conclusion and discussion are a little weak at the moment. It is suggested that at the end, it can help the reader to summarize the current type of Chatbots horizontally and the feasible pedagogical uses. In addition, suggestions for future chatbot development and application in teaching can be made.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Dear Authors,

After a very pleasant reading of the article entitled A Review of AI-Driven Conversational Chatbots Implementation Methodologies and Challenges (1999-2022) I am in favor of publishing it in the Journal Sustainability.

This article makes a very current bibliographic review on the construction of Chatbot systems using machine learning techniques. It is interesting to note that this bibliographic review was based on three fundamental points: (i) objectives of building, (ii) methods and datasets, and (iii) outcomes and challengers. After a thorough SCOPUS search, the authors selected the 28 most important and most current articles on chatbots and developed their article around this.

Also, here are some suggestions for improvements:

1. (p. 1, first paragraph). Where it says “LTSM” would not it be “LSTM”? The meaning of this acronym would be “Long Short-Term Memory”, right?

2. (pg. 1) I think it's prudent where it reads “RNN, Seq2Seq, LTSM, BERT, GPT” to put the full name of the acronyms RNN, LSTM, BERT, and GPT. Not everyone who will read your article will be an expert in machine learning.

3. This is just a suggestion: Use the term “Artificial Neural Networks” in the abstract or in the Introduction, given their importance, currently, in this topic. For example, in the abstract, replace “..., such as natural language processing (NLP) …” with “..., such as natural language processing (NLP) and artificial neural network …”

4. End “Section 1” with a brief paragraph saying what will be accomplished in Sections 2, 3, 4, and 5.

5. (p. 3, First Paragraph of Section 3.1) Replace “1998 to 2022” with “1999 to 2022” or change “1999” in the article title to “1998”.

6. It is good formatting practice to have the entire Table printed on a single page. For example, arrange this in Tables 5 and 6.

Congratulations to the authors for the good work done on this article. I learned many interesting things from them.

Yours sincerely,

The Reviewer

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 4 Report

Strengths:

The review attempts to cover a wide period of time.

Weaknesses:

The paper needs grammar checks and proof-reading to correct things, for example lines 13, 38, 127, and more. I think because of the grammar issues, the intended meaning of the authors in some of the sentences is not what is conveyed in them.

Line 35 mixes up NLP models/architectures with reinforcement learning (a type of ML technique). This should be corrected.

For a literature review paper, I think the number of citations is too small at 28, six of which are preprints. Peer-reviewed versions of the preprints should be cited instead if they exist.

The literature review section does not mention any previous literature review in this field.

Line 122 says the review period is from 1998, which conflicts with the title (1999).

Figure 2 RQ1 conflicts with line 156-158, as there are 4 different points instead of 3. Furthermore, the objectives lack clear reasons for dividing them as such. In addition, “performance improvement” as an objective of building a conversational system is very debatable because the objective is usually tied to a specific task.

Table 3 mixes up again architectures and techniques. This is confusing.

BERT is mentioned a few times in the paper but the original paper by Devlin et al doesn’t seem to be referenced at all.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf