This paper presents an attempt to study the aesthetics of khayal music with reference to the flexibility exercised by artists in performing well-known compositions. We study expressive timing and pitch variations of the given lyrical content within and across performances and propose computational representations that can discriminate between performances of the same song in terms of expression. We employ a dataset of two songs in two ragas each rendered by several prominent artists.

Q2 ( I am an expert on the topic of the paper.)

Disagree

Q3 ( The title and abstract reflect the content of the paper.)

Agree

Q4 (The paper discusses, cites and compares with all relevant related work.)

Agree

Q6 (Readability and paper organization: The writing and language are clear and structured in a logical manner.)

Agree

Q7 (The paper adheres to ISMIR 2025 submission guidelines (uses the ISMIR 2025 template, has at most 6 pages of technical content followed by “n” pages of references or ethical considerations, references are well formatted). If you selected “No”, please explain the issue in your comments.)

Yes

Q8 (Relevance of the topic to ISMIR: The topic of the paper is relevant to the ISMIR community. Note that submissions of novel music-related topics, tasks, and applications are highly encouraged. If you think that the paper has merit but does not exactly match the topics of ISMIR, please do not simply reject the paper but instead communicate this to the Program Committee Chairs. Please do not penalize the paper when the proposed method can also be applied to non-music domains if it is shown to be useful in music domains.)

Agree

Q9 (Scholarly/scientific quality: The content is scientifically correct.)

Agree

Q11 (Novelty of the paper: The paper provides novel methods, applications, findings or results. Please do not narrowly view "novelty" as only new methods or theories. Papers proposing novel musical applications of existing methods from other research fields are considered novel at ISMIR conferences.)

Agree

Q12 (The paper provides all the necessary details or material to reproduce the results described in the paper. Keep in mind that ISMIR respects the diversity of academic disciplines, backgrounds, and approaches. Although ISMIR has a tradition of publishing open datasets and open-source projects to enhance the scientific reproducibility, ISMIR accepts submissions using proprietary datasets and implementations that are not sharable. Please do not simply reject the paper when proprietary datasets or implementations are used.)

Disagree

Q13 (Pioneering proposals: This paper proposes a novel topic, task or application. Since this is intended to encourage brave new ideas and challenges, papers rated “Strongly Agree” and “Agree” can be highlighted, but please do not penalize papers rated “Disagree” or “Strongly Disagree”. Keep in mind that it is often difficult to provide baseline comparisons for novel topics, tasks, or applications. If you think that the novelty is high but the evaluation is weak, please do not simply reject the paper but carefully assess the value of the paper for the community.)

Disagree (Standard topic, task, or application)

Q14 (Reusable insights: The paper provides reusable insights (i.e. the capacity to gain an accurate and deep understanding). Such insights may go beyond the scope of the paper, domain or application, in order to build up consistent knowledge across the MIR community.)

Agree

Q15 (Please explain your assessment of reusable insights in the paper.)

I say "agree' because the methods and dataset itself theoretically offer the capacity to gain accurate and deep understanding of both inter and intra-performer performance analysis (e.g., micro pitch and timing deviations from experts to novice performers). While the authors are indeed pressed for time and space in the article, I'm not sure they presented the most insightful material in the paper itself.

Q16 ( Write ONE line (in your own words) with the main take-home message from the paper.)

Through a combination of MIR tools and techniques (stem separation, F0 extraction, onset detection, etc.) the authors assemble a small but useful dataset for analyzing performance practices in Hindustani bandish performances, offering note-level timings, syllables, and F0 estimates.

Q17 (This paper is of award-winning quality.)

No

Q19 (Potential to generate discourse: The paper will generate discourse at the ISMIR conference or have a large influence/impact on the future of the ISMIR community.)

Disagree

Q20 (Overall evaluation (to be completed before the discussion phase): Please first evaluate before the discussion phase. Keep in mind that minor flaws can be corrected, and should not be a reason to reject a paper. Please familiarize yourself with the reviewer guidelines at https://ismir.net/reviewer-guidelines.)

Weak accept

Q21 (Main review and comments for the authors (to be completed before the discussion phase). Please summarize strengths and weaknesses of the paper. It is essential that you justify the reason for the overall evaluation score in detail. Keep in mind that belittling or sarcastic comments are not appropriate.)

This is a lovely paper. Well written and sits right in between music performance analysis and computational (ethno/)musicology. While I think the dataset is indeed useful, it remains rather small (only 2 songs x 10 artists). I believe that the reason it is small is that despite the many automated methods for information retrieval used, there obviously was a lot of manual researcher intervention and clean up (which is time and labor intensive). In addition, the authors do not report the initial error rates for f0 detection nor onset detection. Onset detection for voice, in particular, is notoriously challenging and error prone. In fact, the authors had a rather clever approach to this problem, but we have no idea what the original error rate was (for any methodology, in fact). The authors simply mention two points as justification for this omission (1) "Recently, a small comparative study of two performances of the same bandish,...served as a preliminary validation for methodology using automatically detected onsets..." and (2) "The resulting alignments are checked and corrected for the occasional errors that arise mainly due to the presence of long vowels in singing." My main issue with the paper is the lack of clarity and transparency in the methodology (see review & metareview). For future researchers who may want to take advantage of similar methodologies to more fully automate many of these processes, especially for non-Western music, I encourage the authors to include an evaluation and discussion of the automated methods. This will bring more value to the paper than merely justification of the 'trustworthiness' of manually-cleaned data.

Minor comments: Lines 21-24: This is a very specific and odd "motivation" presented in extremely passive (contorted even) language. Suggest changing to something like "A motivation for this work was to develop a computational representation of bandish performance practice for computational musicology research. Analyzing within and across performers can provide insights into this understudied performance practice."

Lines 34-36: This is an odd end to this paragraph. I think the authors are just trying to introduce the ensuing paragraphs. Suggest removing.

Lines 49-52: Another work investigating ornamentation in alap you may wish to query is that of Jain et al. “An Algorithmic Approach to Automated Symbolic Transcription of Hindustani Vocals” (2023)

Lines 171-172: The "suitable grouping of phones" is hardly a simplistic process for creating words/lyrics. I strongly suggest the authors elaborate here.

Lines 178-179: It would be appreciated if the authors could clarify the signed relationship. I would have thought that -1 would be a lag and +1 would be ahead of the beat. However, based on the language and the plot it seems the authors mean the inverse. If it is the inverse then specifying clearly would certainly be helpful.

Figure 3 caption: I suggest changing "as a fraction of beat duration" to "as measured in units of beat duration" since the amount can and often is a multiple and not only a fraction. (In addition, mentioning somewhere in the text the actual millisecond distance of 1 beat for this particular example might be helpful, since being ahead or behind an entire beat seems quite extreme, but perhaps "beats" in this style is more akin to subdivisions of a beat in Western music.)

Table 1: The "Swar" row wraps because the authors are listing the full collection. I might suggest replacing this with the raga instead, since the set of swara (I believe) are more or less fixed with a given raga.

Lines 219-224: This description or explanation is difficult to follow in consultation with Figure 3, since the prose uses anglicized syllables, while the figure using the native language. (In other words, I have no idea which of those syllables are the 'Man' syllables). I think it is fine to just pick one, but be consistent.

Lines 238-242: I don't understand this, I think because there are some dashes and commas missing. E.g., I think the authors mean "variable-dimension F0 contour" (and is it really dimension that is changing or just length?) But what is meant by assigning a "lower" and "fixed" dimensional vector when it is inherently variable? The word "piece" here may be contributing to the ambiguity, since in Western music this word commonly refers to an entire work, which I don't believe makes any sense in this context. Could the authors kindly clarify?

Lines 246-247: Any rationale for the choice of 10 uniform intervals per beat duration? Seems a bit arbitrary.

Figure 4: This image is too small to read the axes, even blown up. Again, the text ("Ja1" in English does not align with the lyrics that are typed in the original language. Kindly add the original language syllable text to the caption for clarity.) In addition, the figure unnecessarily shows the ensuing 2-3 syllables as well. Why is this? It is unclear what the blue dotted vertical lines represent, nor why in the 3rd subfigure these lines are compressed. Please align all 3 such that the time durations are comparable (I would suggest not using time in seconds but duration of the beats given the tempo such that beats align; or else do some modest time normalization -- e.g., line 261 talks about time alignment).

Lines 257-263: While Levenshtein distance can be used on vectors as well as strings, the authors do say "strings" here, but it is unclear how the pitch contours became strings when they were clearly numeric. Is there some kind of tying together of the syllable and the pitch value? Some elaboration there would be helpful.

Line 350: I think should say "near the end OF the word or the cycle"

Line 351: The "should" here in this context I presume comes from the authors' domain knowledge? (This would be helpful to clarify).

Q22 (Final recommendation (to be completed after the discussion phase) Please give a final recommendation after the discussion phase. In the final recommendation, please do not simply average the scores of the reviewers. Note that the number of recommendation options for reviewers is different from the number of options here. We encourage you to take a stand, and preferably avoid “weak accepts” or “weak rejects” if possible.)

Weak accept

Q23 (Meta-review and final comments for authors (to be completed after the discussion phase))

This is a well written and fascinating paper, but the details of the methodology are at times unclear or ambiguous, making the overall conclusions of the paper rather weak. For example, the main way the authors compare expressiveness in the pitch differences across syllable instances is "by computing the similarity of syllable pitch contours for pairs drawn from the set of repetitions...[using] Levenshtein distance between string..." However, this is the last paragraph of the Methods ("Measuring Expressiveness") section with no elaboration at all. It is unclear how a syllable marked over a region of f0 contours becomes a string character to perform LD on, nor which LD method they used, since of course the strings would be different overall lengths. I presume that the methodology (while still slightly unclear) is hinted at 244-254 where the authors imply that they are slicing the duration of a single syllable as a function of 10ths of a beat. I would presume, then if the full beat rested on the swar "Sa" that you would have 10 strings of "S" in a row? But this is never explained explicitly in connection to the LD similarity scoring, making steps of the methodology unclear. In addition, there was no mention of the binning of the pitch and how the frequency was mapped to the swara (the authors imply using cents but don't mention the intonation. Other papers have used just tuning for this and not equal temperament). The authors mention time alignment 'optimal time alignment' however this is also non-trivial; are they aligned to the start? To the peak? Etc. More detail is needed in order to be able to understand and potentially replicate this methodology. I strongly encourage the authors to add as much detail as possible should this work be accepted for publication by the committee.

Reviewers (especially myself) have left suggestions for edits to improve the paper. We thank you for your contribution!

Q2 ( I am an expert on the topic of the paper.)

Strongly disagree

Q3 (The title and abstract reflect the content of the paper.)

Agree

Q4 (The paper discusses, cites and compares with all relevant related work)

Agree

Q6 (Readability and paper organization: The writing and language are clear and structured in a logical manner.)

Agree

Q7 (The paper adheres to ISMIR 2025 submission guidelines (uses the ISMIR 2025 template, has at most 6 pages of technical content followed by “n” pages of references or ethical considerations, references are well formatted). If you selected “No”, please explain the issue in your comments.)

Yes

Q8 (Relevance of the topic to ISMIR: The topic of the paper is relevant to the ISMIR community. Note that submissions of novel music-related topics, tasks, and applications are highly encouraged. If you think that the paper has merit but does not exactly match the topics of ISMIR, please do not simply reject the paper but instead communicate this to the Program Committee Chairs. Please do not penalize the paper when the proposed method can also be applied to non-music domains if it is shown to be useful in music domains.)

Agree

Q9 (Scholarly/scientific quality: The content is scientifically correct.)

Agree

Q11 (Novelty of the paper: The paper provides novel methods, applications, findings or results. Please do not narrowly view "novelty" as only new methods or theories. Papers proposing novel musical applications of existing methods from other research fields are considered novel at ISMIR conferences.)

Agree

Q12 (The paper provides all the necessary details or material to reproduce the results described in the paper. Keep in mind that ISMIR respects the diversity of academic disciplines, backgrounds, and approaches. Although ISMIR has a tradition of publishing open datasets and open-source projects to enhance the scientific reproducibility, ISMIR accepts submissions using proprietary datasets and implementations that are not sharable. Please do not simply reject the paper when proprietary datasets or implementations are used.)

Agree

Q13 (Pioneering proposals: This paper proposes a novel topic, task or application. Since this is intended to encourage brave new ideas and challenges, papers rated "Strongly Agree" and "Agree" can be highlighted, but please do not penalize papers rated "Disagree" or "Strongly Disagree". Keep in mind that it is often difficult to provide baseline comparisons for novel topics, tasks, or applications. If you think that the novelty is high but the evaluation is weak, please do not simply reject the paper but carefully assess the value of the paper for the community.)

Agree (Novel topic, task, or application)

Q14 (Reusable insights: The paper provides reusable insights (i.e. the capacity to gain an accurate and deep understanding). Such insights may go beyond the scope of the paper, domain or application, in order to build up consistent knowledge across the MIR community.)

Agree

Q15 (Please explain your assessment of reusable insights in the paper.)

Different expressive variations have systematic patterns in this chosen raga.

Q16 (Write ONE line (in your own words) with the main take-home message from the paper.)

Expressive gestures, particularly in timing and pitch, are more prominent at the beginning and sometimes the end of a bandish line.

Q17 (Would you recommend this paper for an award?)

No

Q19 (Potential to generate discourse: The paper will generate discourse at the ISMIR conference or have a large influence/impact on the future of the ISMIR community.)

Agree

Q20 (Overall evaluation: Keep in mind that minor flaws can be corrected, and should not be a reason to reject a paper. Please familiarize yourself with the reviewer guidelines at https://ismir.net/reviewer-guidelines)

Weak accept

Q21 (Main review and comments for the authors. Please summarize strengths and weaknesses of the paper. It is essential that you justify the reason for the overall evaluation score in detail. Keep in mind that belittling or sarcastic comments are not appropriate.)

The paper investigates how artists inject expressiveness in music performances, and proposes computational methods to identify and analyze these variations within and across performances of the same piece. The study found that expressive gestures, particularly in timing and pitch, are more prominent at the beginning and sometimes the end, with less variation occurring near the boundary.

Strengths The study utilizes a significantly larger dataset compared to previous computational studies on this topic.

It introduces the analysis of pitch-based expressiveness in addition to timing variations.

The paper proposes computational measures capable of discriminating performances based on expressiveness.

Weaknesses The analysis presented in the paper focuses primarily a small part of a bandish.

The chosen modes of variation may not capture all ways a performed could inject expression.

The methods do no work across a piece and need performers to have performed one same section of a piece

Q2 ( I am an expert on the topic of the paper.)

Agree

Q3 (The title and abstract reflect the content of the paper.)

Strongly disagree

Q4 (The paper discusses, cites and compares with all relevant related work)

Disagree

Q5 (Please justify the previous choice (Required if “Strongly Disagree” or “Disagree” is chosen, otherwise write "n/a"))

Two references are needed and must be compared for similarity of the problem and the technique used.

1. Sankaran, Sridharan, P. V. Krishnaraj Sekhar, and A. Murthy Hema. "Automatic segmentation of composition in carnatic music using time-frequency cfcc templates." Proceedings of 11th international symposium on computer music multidisciplinary research. 2015. This paper on Carnatic music tracks sangatis, which are lineage-dependent variations of the same lyric lines, just like in a bandish.

2. VS Viraraghavan, R Gavas, H Murthy, R Aravind, "Visualizing carnatic music as projectile motion in a uniform gravitational field," - Proc. Workshop on Speech, Music and Mind 2019, 2019. This paper observes in Carnatic music the non-uniform variation in pitch components across speed and is relevant to Figure 4 and Section 4.2

Q6 (Readability and paper organization: The writing and language are clear and structured in a logical manner.)

Strongly agree

Q7 (The paper adheres to ISMIR 2025 submission guidelines (uses the ISMIR 2025 template, has at most 6 pages of technical content followed by “n” pages of references or ethical considerations, references are well formatted). If you selected “No”, please explain the issue in your comments.)

Yes

Q8 (Relevance of the topic to ISMIR: The topic of the paper is relevant to the ISMIR community. Note that submissions of novel music-related topics, tasks, and applications are highly encouraged. If you think that the paper has merit but does not exactly match the topics of ISMIR, please do not simply reject the paper but instead communicate this to the Program Committee Chairs. Please do not penalize the paper when the proposed method can also be applied to non-music domains if it is shown to be useful in music domains.)

Strongly agree

Q9 (Scholarly/scientific quality: The content is scientifically correct.)

Agree

Q11 (Novelty of the paper: The paper provides novel methods, applications, findings or results. Please do not narrowly view "novelty" as only new methods or theories. Papers proposing novel musical applications of existing methods from other research fields are considered novel at ISMIR conferences.)

Strongly agree

Q12 (The paper provides all the necessary details or material to reproduce the results described in the paper. Keep in mind that ISMIR respects the diversity of academic disciplines, backgrounds, and approaches. Although ISMIR has a tradition of publishing open datasets and open-source projects to enhance the scientific reproducibility, ISMIR accepts submissions using proprietary datasets and implementations that are not sharable. Please do not simply reject the paper when proprietary datasets or implementations are used.)

Agree

Q13 (Pioneering proposals: This paper proposes a novel topic, task or application. Since this is intended to encourage brave new ideas and challenges, papers rated "Strongly Agree" and "Agree" can be highlighted, but please do not penalize papers rated "Disagree" or "Strongly Disagree". Keep in mind that it is often difficult to provide baseline comparisons for novel topics, tasks, or applications. If you think that the novelty is high but the evaluation is weak, please do not simply reject the paper but carefully assess the value of the paper for the community.)

Agree (Novel topic, task, or application)

Q14 (Reusable insights: The paper provides reusable insights (i.e. the capacity to gain an accurate and deep understanding). Such insights may go beyond the scope of the paper, domain or application, in order to build up consistent knowledge across the MIR community.)

Strongly agree

Q15 (Please explain your assessment of reusable insights in the paper.)

The nature of Hindustani (and Carnatic) music is explored for what it is, without recourse to equivalent Western music concepts. The kind of variation explored is so typical of the music genres that it must be studied.

Q16 (Write ONE line (in your own words) with the main take-home message from the paper.)

This paper has a lucid explanation of the nature of variations in Hindustani music (and applies to Carnatic music, in fact).

Q17 (Would you recommend this paper for an award?)

Yes

Q18 ( If yes, please explain why it should be awarded.)

The paper is a pleasure to read. It succinctly captures of the nature of expressiveness (variations) in Hindustani music. It appears to be the first time the problem is laid bare with supporting data rather than in subjective language.

Q19 (Potential to generate discourse: The paper will generate discourse at the ISMIR conference or have a large influence/impact on the future of the ISMIR community.)

Strongly agree

Q20 (Overall evaluation: Keep in mind that minor flaws can be corrected, and should not be a reason to reject a paper. Please familiarize yourself with the reviewer guidelines at https://ismir.net/reviewer-guidelines)

Strong accept

Q21 (Main review and comments for the authors. Please summarize strengths and weaknesses of the paper. It is essential that you justify the reason for the overall evaluation score in detail. Keep in mind that belittling or sarcastic comments are not appropriate.)

Section 2: The summary sentences of the first paragraph mention "ornamentation", but Figure 1 does not seem to have it. Perhaps another smaller figure with such an example will help.

Section 2: The second paragraph can add a short note that these variations are akin to Sangatis in Carnatic music and compare with Sridharan 2015 (see Lit Survey comments). Section 2: Please mention whether these variations are completely extempore, somewhat practiced, or completely crystallized within a lineage. Section 2: At first use of lineage, please add "(gharana, for those familiar with Indian music)"

Section 3: Please mention the percentage of "the occasional errors that arise mainly due to the presence of long vowels" Section 3: "We observe from Figure 2 how the realised onsets lag the canonical locations most of the time." Very interesting!

Section 4.2: Looking at Figure 4, it may be a good idea to identify only more or less constant pitch segments first (Viraraghavan 2019 in Lit. Survey Section) and then find the PAA in between. That way, redundant PAAs (2nd, 4th, 10th, 12th values), which are really forced to a PAA by the algorithm, can be eliminated. This is a suggestion that can be added for future work. No need to change the data in the paper.

Section 4.2: Last paragraph should define the "optimal time alignment" and quantify the "minor temporal shifts".

Editorial Figures 9 and 10 should be together together, but both should appear and be referred before Figures 5 to 8.

Section 5: The first paragraph needs is starting with within-artist variation, but in the next sentence mentions "across artists". Please rewrite with a little more clarity and in the context of the comment regarding placement of Figures 9 and 10.

Q2 ( I am an expert on the topic of the paper.)

Agree

Q3 (The title and abstract reflect the content of the paper.)

Agree

Q4 (The paper discusses, cites and compares with all relevant related work)

Agree

Q6 (Readability and paper organization: The writing and language are clear and structured in a logical manner.)

Agree

Q7 (The paper adheres to ISMIR 2025 submission guidelines (uses the ISMIR 2025 template, has at most 6 pages of technical content followed by “n” pages of references or ethical considerations, references are well formatted). If you selected “No”, please explain the issue in your comments.)

Yes

Q8 (Relevance of the topic to ISMIR: The topic of the paper is relevant to the ISMIR community. Note that submissions of novel music-related topics, tasks, and applications are highly encouraged. If you think that the paper has merit but does not exactly match the topics of ISMIR, please do not simply reject the paper but instead communicate this to the Program Committee Chairs. Please do not penalize the paper when the proposed method can also be applied to non-music domains if it is shown to be useful in music domains.)

Agree

Q9 (Scholarly/scientific quality: The content is scientifically correct.)

Agree

Q11 (Novelty of the paper: The paper provides novel methods, applications, findings or results. Please do not narrowly view "novelty" as only new methods or theories. Papers proposing novel musical applications of existing methods from other research fields are considered novel at ISMIR conferences.)

disagree

Q12 (The paper provides all the necessary details or material to reproduce the results described in the paper. Keep in mind that ISMIR respects the diversity of academic disciplines, backgrounds, and approaches. Although ISMIR has a tradition of publishing open datasets and open-source projects to enhance the scientific reproducibility, ISMIR accepts submissions using proprietary datasets and implementations that are not sharable. Please do not simply reject the paper when proprietary datasets or implementations are used.)

Agree

Q13 (Pioneering proposals: This paper proposes a novel topic, task or application. Since this is intended to encourage brave new ideas and challenges, papers rated "Strongly Agree" and "Agree" can be highlighted, but please do not penalize papers rated "Disagree" or "Strongly Disagree". Keep in mind that it is often difficult to provide baseline comparisons for novel topics, tasks, or applications. If you think that the novelty is high but the evaluation is weak, please do not simply reject the paper but carefully assess the value of the paper for the community.)

Disagree (Standard topic, task, or application)

Q14 (Reusable insights: The paper provides reusable insights (i.e. the capacity to gain an accurate and deep understanding). Such insights may go beyond the scope of the paper, domain or application, in order to build up consistent knowledge across the MIR community.)

Agree

Q15 (Please explain your assessment of reusable insights in the paper.)

There are several tools like tempo arcs that can be used not only in hindustani music, but also in other styles like Candombe drumming, Jazz and flamenco to introduce expressivity.

The idea of correlating melody and rhythm for analysis is very insightful and re-usable.

Q16 (Write ONE line (in your own words) with the main take-home message from the paper.)

The paper explores how hindustani musicians use expressive syllable and pitch timing to bring expressivity to the performance. Two bandishes from two ragas across several artists are analysed for melodic and rhythmic and melodic expressiveness, and enhanced expressiveness is identified.

Q17 (Would you recommend this paper for an award?)

No

Q19 (Potential to generate discourse: The paper will generate discourse at the ISMIR conference or have a large influence/impact on the future of the ISMIR community.)

Agree

Q20 (Overall evaluation: Keep in mind that minor flaws can be corrected, and should not be a reason to reject a paper. Please familiarize yourself with the reviewer guidelines at https://ismir.net/reviewer-guidelines)

Weak accept

Q21 (Main review and comments for the authors. Please summarize strengths and weaknesses of the paper. It is essential that you justify the reason for the overall evaluation score in detail. Keep in mind that belittling or sarcastic comments are not appropriate.)

The work is an extension of the cited work https://arxiv.org/abs/2503.21142

Strengths of the paper: The paper uses the timing analysis from the previous work, but adds a newer dimension to the work by introducing melodic expressivity by analysing how pitch and rhythm contribute to expressivity.

The findings paves the way to more computational musicology on improvisational aspects.

Weaknesses: The introduced dataset is limited to 2 ragas, Yaman and Bhimpalasi out of which Yaman was already part of the previous work.

Both these bandishes dont have variability in tala (Both of them are in Teental in madhya and drut laya). Analysing different taals could give better insight on expressivity.