Examining the sentiment and emotional differences in product and service reviews: The moderating role of culture

Vinh Truong (RMIT University)

REVIEW 5 major objections 6 minor 10 references

Examining the sentiment and emotional differences in product and service reviews: The moderating role of culture

T0 review · 5 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Service reviews carry stronger and more complex emotions than product reviews, and the gap is widest for Eastern consumers.

desk verdict A plausible but under-supported empirical claim about product vs service review sentiment and culture; the dataset labeling and culture coding need to be fixed before the findings can be trusted. read the letter →

arxiv 2507.21057 v1 pith:EQN226GM submitted 2025-05-07 cs.CY

Vinh Truong (RMIT University) This is my paper · ORCID

classification cs.CY

keywords onlinereviewssentimentanalysisproductversusserviceculturalmoderationemotionalexpressionvalence-arousal-dominancee-commerceconsumeremotions

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that reviews of services and products are emotionally different in a systematic way, and that the customer's cultural background changes how that difference shows up. Analyzing six review datasets with machine-learning emotion classifiers, it finds that service reviews score higher in sentiment, arousal, and complex emotion categories such as love, excitement, and disappointment, while product reviews cluster around relief, pride, and caring. Culture moderates the pattern: Eastern reviewers show a much larger rise in sentiment when moving from products to services, while Western reviewers stay flatter or decline. Gender shows no meaningful moderating effect. If the pattern holds, it would mean that product and service reviews are distinct emotional genres, and that review analytics and marketing strategies should be split by category and culture rather than treated as one uniform signal.

What carries the argument

The machinery is a layered emotion-measurement stack applied to each review. Sentiment is scored by a multilingual sentiment model and assigned to positive/negative/neutral buckets by a rule-based sentiment analyzer; the valence–arousal–dominance (VAD) model maps each review onto three continuous emotional dimensions; a six-category basic-emotion classifier and a 27-category complex-emotion classifier supply discrete emotional labels. The statistical core is a multivariate analysis of variance with category as the independent variable and culture and gender as moderators, followed by a moderation regression with the same interaction terms. The conceptual pivot is the product/service dichotomy viewed through a hierarchy of needs: products satisfy basic functional needs while services satisfy relational and self-fulfillment needs, and this difference is what predicts the emotion gap and its cultural moderation.

What would settle it

Rerun the same analysis after relabeling the dataset that came from a service-review platform as a service dataset instead of a product dataset; if the service-over-product sentiment gap and the East–West interaction shrink, flatten, or flip, the central claim depends on the dataset labels rather than on the nature of products and services.

Watch

Extended reading notes

Core claim

The central discovery is that the product/service distinction predicts the emotional register of a review. In the paper's data, services elicit higher sentiment scores, higher arousal, higher dominance scores, and a greater share of complex emotion categories—love, excitement, amusement, optimism, and admiration, but also disgust, disappointment, disapproval, sadness, confusion, and embarrassment—whereas products elicit more relief, pride, and caring. The category effect on sentiment, sentiment category, valence, and dominance is itself moderated by culture: Eastern consumers show a pronounced increase in sentiment when moving from products to services, while Western consumers show little change or a decline. The paper interprets these patterns through a hierarchy-of-needs lens, arguing that services engage higher-order social and self-fulfillment needs and therefore produce stronger and more varied emotions, with individualism–collectivism shaping which needs dominate expression. Gender does not moderate the effect.

Load-bearing premise

The analysis takes the product/service label attached to each of the six datasets as correct even though one dataset was drawn from a service-review platform and labeled as product reviews, and it collapses reviewer nationality into a binary Eastern/Western split; if either of those assumptions is wrong, the category comparisons and the cultural moderation built on them do not hold.

Editorial extensions

If this is right

Treating review sentiment as a single pool will systematically overstate the positivity of service reviews relative to product reviews in any ranking or dashboard.
Service businesses should design around emotional and relational touchpoints, since service reviews carry a wider and more intense emotional spectrum.
In markets with many Eastern consumers, service-heavy campaigns should emphasize experiential and relational value because the sentiment premium of services over products is larger there.
Gender-based personalization of review interpretation is unlikely to pay off; the paper finds no gender moderation of the category effect.
Sentiment classification into positive/neutral/negative labels misses much of the category difference, so analytics that rely only on polarity will understate the service effect.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A natural next test would be to score bundled offerings—products sold with installation, subscription, or customer-support components—and check whether their emotion profiles fall between pure products and pure services.
The East–West moderation could partly reflect platform-specific writing norms rather than deep cultural values; comparing identical products sold through local and international platforms would separate the two.
Using continuous individualism scores per reviewer country instead of a binary East–West split would sharpen or weaken the moderation claim, since the binary coding discards within-group cultural variance.
Because service reviews carry more negative complex emotions like disgust and disappointment, polarity-only dashboards will miss the most actionable warnings hidden in service feedback.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The paper's headline finding—services elicit stronger sentiment than products, especially for Eastern consumers—rests on two unvalidated coding decisions. The ASOS TrustPilot dataset is labeled 'Product' in Table 1 despite Section 3.2 describing it as capturing both products and services, and on a platform like TrustPilot most reviews concern service attributes like delivery and returns. No rule is given for mapping reviewer nationality or location to the Eastern/Western split, so the central moderating variable is not reproducible. The moderation table reports df=2 for Category*Culture even though the paper describes only two cultures; that is either an unreported third level or a misreported analysis. These are load-bearing flaws, not cosmetic ones.

Credit where it is due: the study applies a reasonably broad emotion toolkit (VAD, VADER, GoEmotions) to six datasets, and the basic question—whether product and service reviews differ emotionally—is worth asking. The gender moderation analysis is a sensible addition, and the complex emotion figure is a nice visual. The paper also reports enough statistical detail to see what was tested, which is more than many applied papers do.

The problems, however, run deep. Dataset labels come from the author's judgment with no validation. The culture coding is opaque. No sample sizes, effect sizes, or confidence intervals are reported anywhere. Significance stars are inconsistent: p=.041 gets the same triple asterisk as p<.001. H1 is only partially supported—category is not significant for sentiment category or basic emotion category—yet the abstract and summary claim broad differences. The frequent self-citations are not disqualifying, but they do not strengthen the evidence.

This is a practice-oriented marketing paper. A reader wanting a quick descriptive look at review language might skim it, but the analytic foundation is too shaky for the results to be cited as evidence. The topic is real and the tools are current, so with data, code, and corrected labels it could be reassessed.

If this came across my desk, I would not desk-reject it outright—the flaws are specific and potentially fixable. But I would send it to a referee with instructions to focus on dataset labeling, culture coding, and the reporting inconsistencies. As is, it does not support its claims.

Referee Report

5 major / 6 minor

Summary. The paper examines whether product reviews differ from service reviews in sentiment and emotion (valence, arousal, dominance, basic and advanced emotions), and whether culture and gender moderate these differences. Using six publicly available datasets labeled as product or service, the author applies four pre-trained NLP models to score reviews, then tests hypotheses with MANOVA and a PROCESS moderation analysis. Results show significant category effects on sentiment score, valence, arousal, dominance, and advanced emotion categories, but not on sentiment category or basic emotion categories; culture moderates several outcomes and gender does not. The paper concludes that services evoke stronger and more complex emotional responses and that this gap is especially pronounced for Eastern consumers, extending Maslow's hierarchy and Hofstede's framework.

Significance. The study's ambition is timely: integrating VAD dimensions and fine-grained emotion categories into a comparison of product vs. service reviews, with cultural moderation, addresses a real gap in the e-commerce sentiment literature. The use of publicly available datasets and pre-trained transformer models is reproducible in principle, and the explicit H1-H3 structure makes the empirical claims clear. However, the empirical foundation is not yet reliable: the dataset category labels are inconsistent with the text, culture is not operationalized, and one of the key tables reports impossible degrees of freedom. If the reported effects survive corrected labeling and transparent coding, the findings would be useful for emotion-aware and culture-aware review analytics.

major comments (5)

[Section 3.2, Table 1] The product/service label for ASOS TrustPilot is contradicted by the paper's own description: Section 3.2 states the ASOS TrustPilot Customer Review Dataset 'captures customer experiences with ASOS products and services, focusing on fit, style preferences, and delivery,' yet Table 1 classifies it as 'Product.' TrustPilot is a platform on which reviewers rate the overall service quality of companies, so labeling it a pure product dataset is unjustified. Because every hypothesis test in Section 4 uses Category as the independent variable, this mislabeling can bias the category contrasts, the cultural interactions, and the gender interactions. Please validate the category assignment (e.g., manual coding of a random sample or an explicit rule based on the review text) and rerun the analyses, or provide evidence that the ASOS reviews are predominantly product-focused.
[Section 3.2 and Section 3.4] The manuscript never specifies how an individual review was assigned to 'Eastern' or 'Western' culture. The datasets in Table 1 include Amazon Reviews (which are global) and Celsius Network, and no nationality or location field is described for these; the paper also does not state how Hofstede's dimensions were mapped onto the binary culture variable. Without a coding rule, the Category × Culture interactions in Table 3 and the plots in Figures 3-5 are not reproducible, and the cultural moderation effect cannot be distinguished from dataset-specific confounds. Please provide the exact mapping (e.g., reviewer self-reported country, IP-based location, or platform-specific fields) and report the number of reviews per culture and per category.
[Table 3] All Category × Culture rows in Table 3 report df = 2. With two levels of Category (Product, Service) and two levels of Culture (Eastern, Western), the interaction degrees of freedom should be (2−1)×(2−1) = 1. The reported df = 2 implies an unreported third level of Culture (or of Category), or a mis-specified model. Since the text describes only a binary Eastern/Western split, the moderation results cannot be interpreted as reported. Please clarify the actual number of culture levels and correct the table; if a three-level culture variable was used, the hypotheses and the interpretation of the cultural moderation must be revised accordingly.
[Abstract, Section 4.1, Table 2, Table 4] The support for H1 is overstated. Table 2 shows that Category has no significant effect on Sentiment Category (p = .614) or Basic Emotion Category (p = .621), and Table 3 shows non-significant Category × Culture interactions for Arousal (p = .799) and Advanced Emotion Categories (p = .401). Yet the abstract claims 'clear differences in emotional expression and sentiment between the two' and Section 4.1 concludes that 'services tend to elicit a higher proportion of complex emotions compared to products.' Additionally, Table 4 appears to mark Basic Emotion Category as 'Yes' for H1 despite the non-significant result in Table 2. The conclusions should be restricted to the dependent variables that actually show significant effects, and the discrepancy between Tables 2 and 4 should be corrected.
[Section 3.4 and Section 4] No sample sizes, group means, standard deviations, effect sizes, or confidence intervals are reported. Section 3.4 asserts that 'the dataset, consisting of a relatively equal representation of product and service reviews, male and female consumers, as well as Western and Eastern reviews, enabled balanced comparisons,' but no counts are provided. Without N and effect sizes, the non-significant gender moderation results (Table 3) are uninterpretable—they could reflect a genuinely absent effect or a lack of statistical power—and the magnitude of the claimed cultural effects cannot be evaluated. Please report per-cell sample sizes, means, and either partial η² or Cohen's d with confidence intervals.

minor comments (6)

[Section 4.3.1] The heading is numbered '4..3.1.' (typo).
[Section 2.1] The text contains 'Hofstede’s’s cultural dimensions' and 'the circumplex model of the effect' (should be 'affect').
[Figure 2] Figure 2 is difficult to read with 29 emotion categories plotted against an unlabeled x-axis; consider ordering by the product-service difference and adding data labels.
[Section 3.3] The four Hugging Face models are named but no repository URLs or version identifiers are given; adding these would aid reproducibility.
[Table 4] Table 4 is badly formatted (e.g., 'Basic Emotion CCategory,' missing column separators) and inconsistent with Table 2; it should be regenerated.
[References] The reference list contains incomplete or malformed entries (e.g., Cortis & Davis 2021; Sudirjo et al. 2023; Truong 2024) and duplicate 'Beyond culture' entries (Hall, 1976a/1976b).

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the reported findings are empirical comparisons of machine-labeled review scores across externally assigned categories, not derivations from fitted parameters or from the authors' prior work.

full rationale

The manuscript contains no equation-level derivation in which a prediction is generated from a fitted parameter, no expression that reduces to its own input by construction, and no load-bearing uniqueness claim imported from the authors' prior work. Sentiment, valence, arousal, dominance, and emotion categories are produced by named external models (e.g., VADER, GoEmotions, word2affect_english, bert-base-uncased-emotion) and then compared across product/service, culture, and gender using MANOVA and Hayes' PROCESS. These comparisons are self-contained empirical analyses of the resulting scores; the outcome variables are not used to construct the category labels, the culture grouping, or the gender grouping. Self-citations appear only as background support for research gaps and practical commentary, and the central hypotheses are supported by the reported statistics rather than by those citations. Potential weaknesses -- such as unvalidated dataset category labels, an ambiguous Eastern/Western coding rule, and the reported df=2 for a binary culture interaction -- are data-validity and reporting concerns, not circularity, because they concern input classifications rather than a derivation that assumes its own conclusion. No circular step can be exhibited under the required standard of quoting a specific reduction in the paper's own logic.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The paper contributes no free parameters of its own, but its conclusions rest on unverified domain assumptions: dataset labels are correct, pre-trained models measure emotions equally across cultures, and a binary East/West split captures Hofstede's individualism dimension.

assumptions (4)

domain assumption The six selected datasets provide valid product/service labels and demographic metadata.
Section 3.2 assumes category and culture/gender fields are reliable; no validation is reported.
domain assumption Pre-trained multilingual sentiment and emotion models measure emotions equivalently across cultures.
Section 3.3 applies models trained mostly on English data to Eastern and Western reviews without invariance checks; cross-cultural emotion measurement is a known problem.
domain assumption Hofstede's individualism-collectivism is adequately represented by a binary Eastern/Western split.
Section 3.2 and the conceptual model collapse culture into two groups; no Hofstede scores are attached to individual reviewers.
domain assumption MANOVA assumptions (normality, homogeneity of covariance) hold.
Section 3.4 asserts skewness and kurtosis within acceptable limits but reports no values or Levene's tests.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Examining the sentiment and emotional differences in product and service reviews: The moderating role of culture." pith.science (2026). https://pith.science/paper/EQN226GM

@misc{pith2026250721057,
  author       = {Pith},
  title        = {Pith review of: Examining the sentiment and emotional differences in product and service reviews: The moderating role of culture},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EQN226GM}},
  note         = {Machine review of arXiv:2507.21057}
}

read the original abstract

This study explores how emotions and sentiments differ in customer reviews of products and services on e-commerce platforms. Unlike earlier research that treats all reviews uniformly, this study distinguishes between reviews of products, typically fulfilling basic, functional needs, and services, which often cater to experiential and emotional desires. The findings reveal clear differences in emotional expression and sentiment between the two. Product reviews frequently focus on practicality, such as functionality, reliability, and value for money, and are generally more neutral or pragmatic in tone. In contrast, service reviews involve stronger emotional engagement, as services often entail personal interactions and subjective experiences. Customers express a broader spectrum of emotions, such as joy, frustration, or disappointment when reviewing services, as identified using advanced machine learning techniques. Cultural background further influences these patterns. Consumers from collectivist cultures, as defined by Hofstede cultural dimensions, often use more moderated and socially considerate language, reflecting an emphasis on group harmony. Conversely, consumers from individualist cultures tend to offer more direct, emotionally intense feedback. Notably, gender appears to have minimal impact on sentiment variation, reinforcing the idea that the nature of the offering (product vs. service) and cultural context are the dominant factors. Theoretically, the study extends Maslow hierarchy of needs and Hofstede cultural framework to the domain of online reviews, proposing a model that explains how these dimensions shape consumer expression. Practically, the insights offer valuable guidance for businesses looking to optimize their marketing and customer engagement strategies by aligning messaging and service design with customer expectations across product types and cultural backgrounds.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

10 extracted references · 8 canonical work pages

[1]

AlQahtani, A. S. (2021). Product sentiment analysis for amazon reviews. International Journal of Computer Science & Information Technology (IJCSIT) Vol,

work page 2021
[2]

Hutto, C., & Gilbert, E. (2014). Vader: A parsimonious rule-based model for sentiment analysis of social media text. Proceedings of the international AAAI conference on web and social media, Jang, Y . J., & Kim, W. G. (2011). The role of negative emotions in explaining restaurant customer dissatisfaction and behavioral intention. Kanwal, M., Burki, U., Al...

work page 2014
[8]

Hong, Y ., Huang, N., Burtch, G., & Li, C. (2016). Culture, conformity, and emotional suppression in online reviews. Journal of the Association for Information Systems, 17(11),

work page 2016
[13]

Narratives of social cohesion

Au, N., Buhalis, D., & Law, R. (2014). Online complaining behavior in mainland China hotels: The perception of Chinese and non-Chinese customers. International Journal of Hospitality & Tourism Administration, 15(3), 248-274. Barrett, L. F., & Lida, T. (2024). Constructionist Theories of Emotions in Psychology and Neuroscience. In Emotion Theory: The Routl...

arXiv 2014
[43]

Khalili, L., You, Y ., & Bohannon, J. (2022). BabyBear: Cheap inference triage for expensive language models. arXiv preprint arXiv:2205.11747. Kondopoulos, K. D. (2014). Hospitality from Web 1.0 to Web 3.0. In The Routledge Handbook of Hospitality Management (pp. 277-286). Routledge. Li, H., Liu, H., & Zhang, Z. (2020). Online persuasion of review emotion...

arXiv 2022
[145]

Saunders, M. N. K. (2015). Research Methods for Business Students. In P . Lewis & A. Thornhill (Eds.), (7th ed ed.): Harlow, United Kingdom : Pearson Education Limited. Schuckert, M., Liu, X., & Law, R. (2015). Hospitality and tourism online reviews: Recent trends and future directions. Journal of Travel & Tourism Marketing, 32(5), 608-621. Sparks, B. A.,...

work page 2015
[370]

McFarlane, J. M. (2024). Theories of Emotions. Introduction to Psychology. McLeod, S. (2007). Maslow's hierarchy of needs. Simply psychology, 1(1-18). Mesquita, B. (2022). Between us: How cultures create emotions. WW Norton & Company. Meyers-Levy, J., & Loken, B. (2015). Revisiting gender differences: What we know and what lies ahead. Journal of Consumer ...

work page 2024
[476]

P ., Ghasemaghaei, M., & Hassanein, K

Eslami, S. P ., Ghasemaghaei, M., & Hassanein, K. (2022). Understanding consumer engagement in social media: The role of product lifecycle. Decision Support Systems, 162, 113707. Filieri, R. (2015). What makes online reviews helpful? A diagnosticity-adoption framework to explain informational and normative influences in e-WOM. Journal of Business Research...

work page 2022

Show all 10 references

[1161]

Russell, J. A. (2003). Core affect and the psychological construction of emotion. Psychological review, 110(1),

2003
[1270]

Filieri, R., McLeay, F., Tsui, B., & Lin, Z. (2018). Consumer perceptions of information helpfulness and determinants of purchase intention in online consumer reviews of services. Information & management, 55(8), 956-970. Fischer, A. H., & Manstead, A. S. (2000). The relation ...

2018

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.

[1] [1]

AlQahtani, A. S. (2021). Product sentiment analysis for amazon reviews. International Journal of Computer Science & Information Technology (IJCSIT) Vol,

work page 2021

[2] [2]

Hutto, C., & Gilbert, E. (2014). Vader: A parsimonious rule-based model for sentiment analysis of social media text. Proceedings of the international AAAI conference on web and social media, Jang, Y . J., & Kim, W. G. (2011). The role of negative emotions in explaining restaurant customer dissatisfaction and behavioral intention. Kanwal, M., Burki, U., Al...

work page 2014

[3] [8]

Hong, Y ., Huang, N., Burtch, G., & Li, C. (2016). Culture, conformity, and emotional suppression in online reviews. Journal of the Association for Information Systems, 17(11),

work page 2016

[4] [13]

Narratives of social cohesion

Au, N., Buhalis, D., & Law, R. (2014). Online complaining behavior in mainland China hotels: The perception of Chinese and non-Chinese customers. International Journal of Hospitality & Tourism Administration, 15(3), 248-274. Barrett, L. F., & Lida, T. (2024). Constructionist Theories of Emotions in Psychology and Neuroscience. In Emotion Theory: The Routl...

arXiv 2014

[5] [43]

Khalili, L., You, Y ., & Bohannon, J. (2022). BabyBear: Cheap inference triage for expensive language models. arXiv preprint arXiv:2205.11747. Kondopoulos, K. D. (2014). Hospitality from Web 1.0 to Web 3.0. In The Routledge Handbook of Hospitality Management (pp. 277-286). Routledge. Li, H., Liu, H., & Zhang, Z. (2020). Online persuasion of review emotion...

arXiv 2022

[6] [145]

Saunders, M. N. K. (2015). Research Methods for Business Students. In P . Lewis & A. Thornhill (Eds.), (7th ed ed.): Harlow, United Kingdom : Pearson Education Limited. Schuckert, M., Liu, X., & Law, R. (2015). Hospitality and tourism online reviews: Recent trends and future directions. Journal of Travel & Tourism Marketing, 32(5), 608-621. Sparks, B. A.,...

work page 2015

[7] [370]

McFarlane, J. M. (2024). Theories of Emotions. Introduction to Psychology. McLeod, S. (2007). Maslow's hierarchy of needs. Simply psychology, 1(1-18). Mesquita, B. (2022). Between us: How cultures create emotions. WW Norton & Company. Meyers-Levy, J., & Loken, B. (2015). Revisiting gender differences: What we know and what lies ahead. Journal of Consumer ...

work page 2024

[8] [476]

P ., Ghasemaghaei, M., & Hassanein, K

Eslami, S. P ., Ghasemaghaei, M., & Hassanein, K. (2022). Understanding consumer engagement in social media: The role of product lifecycle. Decision Support Systems, 162, 113707. Filieri, R. (2015). What makes online reviews helpful? A diagnosticity-adoption framework to explain informational and normative influences in e-WOM. Journal of Business Research...

work page 2022