The historical narrative of popular music privileges front-stage stars, and music information retrieval (MIR) has largely reproduced this bias. Backstage laborers—arrangers, session musicians, engineers—remain absent from the field's datasets and research questions. Responding to Born's (2020) call for a critical MIR, we argue that encoding historically marginalized music workers into computational databases constitutes an ontological intervention that grants them digital existence in an era where absence from metadata means erasure from history. We introduce the Private Archive of Scanned Music Artifacts (PASMA), a dataset reconstructed by an LLM pipeline from 1,122 Taiwanese vinyl liner notes spanning 1957–1991, and combine Social Network Analysis with MERT embeddings on vocal-separated stems. The collaboration network transformed sharply around 1978, when multitrack recording shifted it from a fragmented, high-modularity ecology to a centralized giant component bridged by elite intermediaries. MERT embeddings indicate that the arrangement traditions defined by these network communities occupy distinct regions in acoustic space. These findings demonstrate that MIR's computational tools, redirected toward the backstage, can decode historical power structures and restore the visibility of the laborers who built the soundscapes of postwar Taiwan.

Q18. Write ONE line (in your own words) with the main take-home message from the paper.

The MIR/computational musicology world must capture, then use, data that reflects the contributions of all those involved in music production.

Q23. 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.

Overall the paper is simultaneously provocative, inspiring, and interesting. Nice mix of methods employed with good effect. Paper makes the argument that we need to look for and then use "behind the scenes" data to improve our musicological understanding of music production beyond "hit" analyses. The incorporation of audio data that reinforces the findings particularly compelling.

Some small issues:

1. Notwithstanding that space is at a premium, Fig. 2 is hard to read. Do not think I have a solution, however.

2. The paper discusses the pre- and post multitrack eras in Taiwan music production (with mid/late 70's being the border). The paper does demonstrate a strong signal around this time. However, what historical evidence is there that there was indeed an industry-wide shift right around this time in recording technology? Multi-track recording wasn't terribly rare in the 60's or 70's.

Q25. Meta-review and final comments for authors (to be completed after the discussion phase)

The reviewers all agreed that this is a very strong paper on a topic that is not often seen at ISMIR. The paper does a great job of bringing together overlooked data and a number of analytic tools to create a picture of the evolution of music production in post-war Taiwan.

The reviewers consider the paper to be right on the cusp of award-winning. They would have liked to have seen more contextualization of Taiwanese music industry, but this is a non-fatal shortcoming that can be addressed in a revision.

Q18. Write ONE line (in your own words) with the main take-home message from the paper.

How MIR tools combined with computational methods from the social sciences can enrich computational analysis of music that uncovers important shifts in music-making, while simultaneously gives voice to the often-overlooked contributors to musical production.

Q23. 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.

This paper provides an excellent example of how MIR tools combined with computational methods from the social sciences can uncover important shifts in music-making, while simultaneously giving voice to often-overlooked contributors to musical production. The authors rightfully acknowledge that "backstage laborers" — arrangers, session musicians, and sound engineers — are largely absent from existing datasets and research agendas. They propose a methodology for analyzing these hidden contributions and demonstrate its value in detecting sharp structural transformations in a musical style over time. Grounded in Georgina Born's seminal 2020 paper calling for a more intense, non-hierarchical, and agonistic collaboration between the humanities and MIR, the authors develop an approach to grant "digital existence" to historically marginalized music workers within computational databases. A central contribution of the paper is the Private Archive of Scanned Music Artifacts (PASMA) dataset, reconstructed from 1,122 vinyl liner notes via an LLM pipeline that combines Social Network Analysis (SNA) with MERT-based audio embeddings on vocal-separated stems. The paper makes a significant contribution to computational ethnomusicology by providing a methodology that embeds important aspects of musical production into the digital representation of music itself, and by demonstrating how these aspects prove relevant, as expected, in analyzing the evolution of a musical style. It not only gives a voice in the digital domain to the backstage workers so essential to music creation, but also shows how this information illuminates the structural transformations that occur over time within a given musical style. Although the social network analysis methodologies fall somewhat outside my area of expertise, they appear to offer valuable tools for tracing the transformations wrought by the "invisible" network of backstage contributors. The proposed pipeline is clearly extensible to other musical styles, and in many respects this paper presents an exemplary model of how MIR tools can reveal important structural patterns of musical evolution when directed at the collaborative networks that quietly shape popular music production. In doing so, it gives voice to oft-ignored communities while also demonstrating how central they are to the music's evolution. However, in contrast to the detailed treatment of the technical and computational strategies, the paper is notably lacking in historical context for the evolution of popular music in postwar Taiwan. This dimension deserves at least a dedicated section or subsection on the paper. Only a single anonymized source is cited as evidence for the period under analysis, and no substantive socio-cultural or historical framing is provided. This gap significantly weakens section 4 where otherwise extremely compelling cross-references between the computational results and historically meaningful moments — such as “the cross-label recording prohibitions and state policies prioritizing Mandarin,“ in the first period (1957-1976) the influence of the American club circuit in cultivating ensemble-playing experience, and the pivotal role of composer-arranger Chen Chih-Yuan in shaping the popular music scene. None of these historically significant factors are supported by adequate bibliographic evidence, even as the computational analysis graphs eloquently point toward their importance. Figure 3, for instance, reveals the existence of four arrangement traditions according to the authors: a pre-multitrack tradition, a Chen Chih-Yuan tradition, an Electone tradition, and a post-Chen era in Taiwanese popular music. However, this is not corroborated or discussed elsewhere in the paper. This constitutes a major weakness of the paper in my view, and one that runs counter to the very spirit of agonistic, non-hierarchical interdisciplinary collaboration that Born's framework advocates.

Q18. Write ONE line (in your own words) with the main take-home message from the paper.

Redirecting MIR tools toward backstage laborers shows that Taiwan's 1978 multitrack transition restructured both the collaboration network and the recorded sound, making invisible music workers computationally visible.

Q23. 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.

1. Summary The paper introduces PASMA, a dataset reconstructed from the liner notes of 1,122 Taiwanese vinyl records via an LLM-assisted pipeline, models the credits as a time-varying collaboration network, and pairs this with MERT embeddings of vocal-separated stems from a 960-track audio corpus. The central finding is a structural break at 1976–1978, coinciding with the adoption of multitrack recording. The paper frames this as an ontological intervention that restores digitally absent backstage laborers to the historical record.

2. Strengths

1. Pioneering research object. This is exactly the kind of novel topic ISMIR's guidelines encourage. The paper does not merely cite Born's critical-MIR program; it operationalizes it into a concrete dataset, method, and historical finding.
2. Dataset contribution. PASMA represents substantial archival labor on multi-script, heterogeneous sources, with a pseudonymized research release planned. The pipeline is described stage by stage and is transferable to other industries.

3. Weaknesses

1. Embedding evidence is visual only. All acoustic claims rest on inspecting UMAP projections, which distort global structure. Quantitative support is needed.
2. Some Typo. "processed by by" in Line 108; "ALBUMS SCANED", "Verifiy" in Fig. 1.

Q18. Write ONE line (in your own words) with the main take-home message from the paper.

The paper presents a data-driven historical investigation on change of backstage labourers in Taiwanese pop music industry in terms of their collaboration networks and sonic impact.

Q23. 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.

Top level summary:

The paper presents a data-driven historical investigation on change of backstage labourers in Taiwanese pop music industry in terms of their collaboration networks and sonic impact. It contributes a novel dataset containing meta-data of contributors and their roles in the production of the songs. It integrates social network analysis with MERT-based audio embedding to pinpoint important figures and time period of the music tradition, and also to reveal the shifting importance of each production role. The paper advocates, and through its actual practice, for “restoring the digital existence of historically invisible backstage labourers.”

Strengths: [Writing] Very mature writing that is clear, logical, and interesting to read. [Topic] It addresses an important but often neglected topic investigating the evolution of collaboration networks for backstage workers and their sonic impact through MIR techniques [Literature review] Very thorough presentation of the background integrating historical realities of the the music industry, humanity discourse (Born’s critique), and MIR approaches. [Dataset, reusable insight] It introduces a dataset (PASMA) that fills a vacuum of unrepresented music contributors. The dataset curation on its own presents an interesting approach integrating LLM to correct OCR errors and convert the data in more structured format. [Findings] Clearly written results, where all the figures and tables meaningfully connect to the overall topic. Plus, the results are discussed in historical context of the musical tradition.
[Model, reusable insight] I find that correlating the social network analysis and acoustic analysis is a interesting, novel, and meaningful contribution of the paper. It is an interesting application of existing CS/MIR techniques that reveals the musical impact of these backstage workers in a data-driven way.

There are no major weaknesses, but I would prefer further clarifications on the following:

For figure 2, it is not entirely clear how the cumulative time points in snapshot are chosen. Is it chosen a-priori based on domain knowledge (line 94 mentioned 1978 but the year 1976 was not mentioned before)? Or automatically segmented based on some graph algorithm?Based on the beginning of the discussion, I gathered that SNA identifies this grouping, but based on which diagnostic metrics? Also, since the author want to show the accumulation of the network through time, won’t it be more readable to apply three different mask to the final plot, so that the position of the same person don’t change?

Historical context (R1, Meta-review). Addressing the concern that multitrack recording was not rare in the 1960s-70s, we no longer claim an "introduction of multitrack recording around 1978." Both instances now read "the industrial adoption and standardization of multitrack studio workflows in the late 1970s," a claim commensurate with our network evidence. The anonymized period source is restored to its full citation (Cheng 2021). Section 5.2 now closes with a pointer to the project website, which hosts a fuller historical account and complete PASMA documentation. The abstract now states the scope explicitly (Taiwan, 1957-1991).

Acoustic claims (R2). We agree that UMAP projections cannot bear quantitative weight. Four claims were downgraded: "independently corroborating" to "visually consistent with"; "confirming that communities carry distinct acoustic signatures" to "indicating that"; "the sonic evidence of his bridging role" to "visually corresponds to"; and "reproduce this boundary... confirms" to "mirror this boundary in the UMAP projection... suggests." A quantitative protocol (PERMANOVA and silhouette analysis with permutation tests on the raw embeddings) is planned for the journal version. The abstract is aligned accordingly.

Snapshot time points (R3). Section 4.1 now states that the 1976 and 1978 endpoints bracket the transition identified by the five-year sliding-window metrics, and that 1991 marks the end of the PASMA study period.

Typos (R2). "by by," "sources.The," and the Figure 1 labels ("ALBUMS SCANED," "Verifiy") are corrected.

Other changes. Acknowledgments added (omitted for anonymous review). Bibliography corrected, including the full author list and venue for the MERT reference (ICLR 2024). Figure captions tightened for length.