Pith. sign in

REVIEW 4 major objections 5 minor 42 references

Bridging Subjective and Objective QoE: Operator-Level Aggregation Using LLM-Based Comment Analysis and Network MOS Comparison

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper claims that LLM-scored live-stream comments, filtered for performance complaints and aggregated per provider in five-minute windows, can detect service degradations through comment trends alone.

desk verdict A genuinely new dataset and aggregation pipeline, but the outage-detection validation is circular and the paper's central claim doesn't survive contact with its own methods. read the letter →

arxiv 2506.00924 v2 pith:MMOR2QQI submitted 2025-06-01 cs.NI cs.AIcs.HC

Parsa Hassani Shariat Panahi , Amir Hossein Jalilvand , M. Hassan Najafi This is my paper
classification cs.NIcs.AIcs.HC
keywords qualityofexperienceQoEmonitoringlive-streamcommentsLLMscoringmeanopinionscoredeltaMOSITU-TP.1203semanticfiltering
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 tries to establish that unstructured viewer comments on live-streaming platforms can be turned into a high-resolution, operator-level quality-of-experience (QoE) monitoring signal. It builds a two-layer framework: an objective layer that predicts video MOS from network key performance indicators using a machine-learning model trained on ITU-T P.1203 reference scores, and a subjective layer that filters comments for performance complaints and uses a large language model to assign each a scalar 1\textendash 5 MOS score. The scores are aggregated by simulated Internet service provider and into five-minute time windows, and a delta MOS metric measures each provider's deviation from the platform-wide average. A controlled simulation in which all of one provider's comment scores are forcibly lowered inside a 90-minute window shows the delta signal dropping below a threshold, which the authors present as evidence that comment-derived trends alone can localize a disruption. If the framework holds, operators would receive privacy-preserving, per-provider subjective MOS reports every five minutes without any client-side instrumentation or video content analysis.

What carries the argument

The central object is the delta MOS metric, defined as $\Delta\text{MOS}_{i,t} = \mu_{i,t} - \text{GlobalMOS}_t$, where $\mu_{i,t}$ is the average LLM-derived MOS for provider $i$ in time window $t$ and $\text{GlobalMOS}_t$ is the equal-weight platform-wide average over providers in the same window. This subtraction is what separates provider-specific shifts from platform-wide trends, so a negative deviation flags a localized problem while a value near zero means the provider tracks the platform baseline. The supporting machinery is the semantic filter (problem-plus-context keyword matching, meme rejection, and embedding similarity against anchors with a higher threshold for short comments), the deterministic LLM scoring at temperature zero, and the aggregation of comments into five-minute windows.

What would settle it

Correlate real, per-provider comment-derived MOS against a known network incident, such as an announced regional ISP outage, using true ISP attribution rather than random assignment; the central claim fails if comment-derived MOS does not drop significantly during the incident for the affected provider while staying stable for unaffected providers.

Watch

Extended reading notes

Core claim

The central discovery is that a two-stage pipeline can convert noisy live-stream chat into a per-operator, time-resolved subjective quality metric that is directly comparable to network-derived MOS. The filtering stage keeps only comments that mention both a problem (such as lag or buffering) and a streaming or network context, rejects meme phrases, and applies embedding-based cosine similarity to anchor phrases with length-sensitive thresholds. The scoring stage prompts an instruction-tuned multilingual LLM in deterministic mode to return a MOS value from 1 to 5, or \textendash 1 for irrelevant text, yielding reproducible scalar scores. Once each comment is randomly assigned to one of three simulated ISPs and given a synthetic timestamp, per-window averages are computed, and subtracting the platform-wide mean from each provider's average defines the delta MOS. The paper argues that a sustained negative delta identifies provider-specific degradation even without network telemetry, demonstrating this with an injected outage in which every one of ISP3's comments inside the window is overwritten to the minimum score.

Load-bearing premise

The outage-detection result depends on the untested premise that real viewer complaints, and the LLM scores they receive, actually move when genuine network quality degrades, since the simulation sidesteps that link by directly overwriting all of one provider's scores inside the outage window.

Editorial extensions

If this is right

  • Each operator would receive, every five minutes, its own subjective MOS and the platform-wide average, letting it distinguish a quality drop confined to its infrastructure from an issue affecting all providers.
  • A per-ISP delta MOS below an operator-configured threshold (the paper illustrates \textendash 0.4) can act as an outage alert driven purely by comment-derived sentiment.
  • Operators can triangulate the comment-based MOS against objective MOS from their own telemetry or from the P.1203-based model, surfacing cases where network KPIs look healthy but users report poor experiences.
  • The pipeline scales because filtering is keyword-plus-embedding based and each comment is scored by a single-turn LLM call, with no per-user data exposure since only windowed aggregates are reported.

Reading between the lines

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

  • The outage simulation validates the aggregation arithmetic but not the sensing chain: because the simulated event overwrites the outcome variable (every comment score becomes 1.0) rather than perturbing network state and letting real comments respond, the experiment cannot by itself show that genuine outages would move comment-derived MOS.
  • A field test would close that gap: real ISP attribution and known outage records could be correlated against comment-derived per-ISP MOS, or LLM scores from a measured degraded window compared against a control window.
  • The uniform random ISP assignment means any baseline per-provider structure in the data is noise; once real attribution is available, the paper's equal-user-base and uniform-comment-volume assumptions would need to be replaced by volume- and engagement-weighted aggregation.
  • The same delta MOS machinery transfers to any time-stamped user text with inferable provider or regional attribution, including support forums, app-store reviews, and multiplayer game chat.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The manuscript proposes a dual-layer QoE monitoring framework: an objective layer that maps network KPIs (latency, jitter, loss, throughput) to MOS via a Random Forest model trained on ITU-T P.1203 reference scores, and a subjective layer that filters live-stream comments for QoE relevance, assigns scalar MOS values with a decoder-only LLM at temperature 0, and aggregates these scores per simulated ISP in 5-minute windows. A new metric, delta MOS (the per-ISP average minus the platform-wide average per window), is introduced, and a 'controlled outage simulation' in Section 6.4 is presented as evidence that comment-based trends alone can detect provider-specific degradations. The dataset consists of 47,894 comments, of which 33,770 receive valid MOS scores after filtering, with synthetic timestamps at 3-second spacing and uniform random ISP labels.

Significance. The framework's mechanics are transparent and re-implementable: Eqs. (1)-(3) correctly define per-provider averaging, platform-wide averaging, and per-provider deviation, and the filtering stages are described in sufficient detail to reproduce. The dataset, with roughly 34,000 QoE-relevant comments, is a potentially useful artifact for future work on comment-based QoE. However, the central validation claim is not supported. The outage experiment in Section 6.4 directly overwrites the outcome variable (MOS_i = 1.0 for all ISP3 comments in the window), so the resulting delta-MOS drop is guaranteed by construction and does not test whether comment text, volume, or timing responds to network degradation. In addition, the random assignment of ISP labels in Sections 4.5 and 5.3 means that all per-ISP structure in the base data is noise, and the LLM-assigned MOS values are never calibrated against human subjective ratings or an ITU-standard subjective dataset.

major comments (4)
  1. [Section 6.4, Eq. (4)] The simulated outage is circular and therefore does not support the abstract's claim that service disruptions are identified 'through comment-based trends alone.' Eq. (4) forcibly sets MOS_i = 1.0 for every ISP3 comment in [15:00, 16:30], and the subsequent re-aggregation then subtracts the global mean from this depressed ISP3 average, guaranteeing a negative delta MOS that crosses the -0.4 threshold. The experiment would pass even if the comments were unrelated to network quality, if the LLM scored randomly, or if the pipeline ignored comment text entirely; it tests only the arithmetic of Eqs. (1)-(3). A valid simulation would need to inject a change in comment content, volume, or timing plausibly associated with an outage and then verify that the pipeline recovers the event, but no such test is provided.
  2. [Sections 4.5 and 5.3] Uniform random assignment of comments to ISP1-ISP3 does not model operator-level attribution and undermines the per-ISP analysis. With random labels, any ISP-specific structure in the base data is statistical noise, and the only deterministic per-provider signal in the entire paper is the hand-injected overwrite of Eq. (4). Consequently, Figs. 6-9 cannot demonstrate that the framework isolates provider-specific phenomena; at most they show the aggregation formulas applied to synthetic metadata. The paper needs either real IP-range-based attribution or a realistic generative model of user-to-ISP association before claims of operator-level detection are justified.
  3. [Sections 4.4 and 5.2] The LLM-assigned scalar MOS values are never validated against human ratings or standardized subjective QoE data. The prompt defines a 5-point scale, and Fig. 5 shows a heavily skewed distribution, but no evidence is given that the scores correspond to user-perceived quality in any measurable sense. Since the entire subjective side, and the later comparison with objective P.1203-based MOS, depends on the semantic validity of these scores, the absence of any calibration study is a load-bearing gap. The paper should compare LLM scores against at least a small human-annotated set, or against established subjective MOS datasets, before claiming that comment-derived MOS can serve as a 'high-resolution QoE monitoring signal.'
  4. [Section 6.1 and Section 5.3] The modeling assumptions of equal user bases and uniform comment volume over time are acknowledged, but the synthetic 3-second timestamp spacing makes Figs. 6-10 a property of simulated metadata rather than of observed comment dynamics. Real comment streams exhibit bursts, pauses, and diurnal variation, and the paper's own stated future work concedes that temporal normalization would be needed in practice. As a result, the time-series visualizations cannot be read as evidence about real-world feedback behavior; they are demonstrations of the aggregation code on an artificial schedule.
minor comments (5)
  1. [Figure 9(c) caption] The caption of Fig. 9(c) reads 'ISP2 MOS Over Time' even though the panel is described as the ISP3 time series, and the caption text repeats the same phrasing as the ISP2 panel; the figure and caption should be aligned to avoid mislabeling.
  2. [Section 6.5] The final sentence of Section 6.5 ends with 'It also supports real-time' and is incomplete; the sentence should be finished or removed.
  3. [Table 1 and Section 5.1] Table 1 has a typo in the header ('Refrence'), and Section 5.1 refers to the field 'commen mos' instead of 'comment mos'; these small errors should be corrected.
  4. [Section 4.4] The description of the LLM is intentionally model-agnostic ('a multilingual, multimodal, instruction-tuned transformer released in 2024'), but the reproducibility claim in Section 2 would be stronger if the exact model identifier, version, and prompt template were provided in the text or in an appendix.
  5. [Abstract and Section 6.4] The phrase 'controlled outage simulation' is misleading because Eq. (4) is a manual overwrite of the output variable rather than a simulation of comment-level effects; rewording to 'synthetic outcome injection' would more accurately describe the experiment.

Circularity Check

2 steps flagged · score 8.0 of 10

The outage-detection demonstration is forced by construction: Eq. (4) overwrites ISP3's MOS to 1.0 inside the test window, and Fig. 10 reads the resulting DeltaMOS drop back out; the filter's problem-keyword requirement also manufactures the low-MOS distribution presented as behavioral evidence.

  1. self definitional [Section 6.4 (Simulated Outage Detection), Eq. (4), and Fig. 10]
    "We inject a synthetic outage into the dataset by manually degrading the MOS for ISP3 within a fixed temporal interval. Specifically, for all user comments associated with ISP3 between 15:00 and 16:30 on January 1, 2024, the MOS value is forcibly set to 1.0... Formally, the synthetic outage is applied using the following mask: ∀i∈ ISP3, if t∈ [15:00, 16:30] =⇒ MOSi = 1.0 (4)... The impact of this simulated outage is visualized in Fig. 10, which shows a sharp and sustained drop in ISP3's ∆MOS during the affected interval."

    The experiment sets the outcome variable (per-comment MOS) for the target ISP inside the window, then re-aggregates via Eq. (1) and subtracts the global mean via Eq. (3). Because ISP3's window average is forced to 1.0 while the other ISPs keep their baseline values, the negative DeltaMOS crossing the -0.4 threshold follows arithmetically; the detection is the injection read back. The test would pass even if comments had no relation to network quality, if the LLM scored randomly, or if the text were ignored. No link between comment content/volume and real degradation is tested.

  2. other [Section 4.3 (Semantic Filtering to Isolate QoE-Relevant Feedback) and Section 5.2 (Subjective Score Distribution), Fig. 5]
    "For a comment to be considered QoE-relevant, it must: Match at least one problem keyword, Match at least one context keyword, Not match any meme keyword... The distribution is heavily skewed toward lower values, with scores of 1 and 2 dominating. This reflects typical user behavior on live-streaming platforms, where negative experiences prompt explicit feedback, while positive or neutral ones are often unreported or ambiguously expressed."

    The inclusion rule requires every retained comment to contain a problem keyword (e.g., 'buffer', 'lagging'), so the retained corpus is, by construction, complaint-dominated. The low-MOS skew in Fig. 5 is therefore a direct consequence of the filter, yet the text presents it as evidence about typical user behavior ('negative experiences prompt explicit feedback') and later affirms 'the practical assumption that subjective QoE signal is louder when quality fails.' The selection rule manufactures the very distribution offered as validation.

full rationale

The central demonstration of Section 6.4 is circular. Eq. (4) does not simulate a change in comment content, volume, or timing; it forcibly assigns MOS_i = 1.0 to every ISP3 comment in [15:00, 16:30]. Re-aggregation (Eq. 1) and subtraction of the global mean (Eq. 3) then guarantee a negative DeltaMOS that crosses the -0.4 threshold, so Fig. 10's 'detection' is the injection read back. The experiment would pass even if comments were unrelated to network quality or if the LLM scored randomly; it tests only the arithmetic of the averaging and thresholding pipeline. A second, milder circularity appears in Sections 4.3 and 5.2: the filter requires a problem keyword, so the low-MOS skew in Fig. 5 is manufactured by the inclusion rule yet reported as evidence of user behavior. The random ISP assignment in Section 5.3 also means per-ISP baseline structure is noise; the only deterministic provider-specific signal is the hand-injected overwrite. Positively, Eqs. (1)-(3), the filtering stages, and the 5-minute aggregation are transparent and re-implementable, and the objective P.1203 model rests on prior work [9] rather than on the present experiment, so the self-citation is not itself circular. The lack of validation of LLM-assigned MOS against human ratings is a correctness risk, not a circularity. Because the headline 'detection' reduces by construction to its own input, the circularity score is 8.

Assumptions & free parameters 9 free parameters · 6 assumptions · 1 invented entities

The ledger holds 9 free parameters, 6 axioms, and 1 invented metric. The costliest items: the LLM-as-MOS validity axiom (no human or ITU ground truth in this domain), the outage-overwrite axiom (Eq. 4 replaces the outcome instead of perturbing a cause), and random ISP assignment, which strips the operator-level axis of real signal. The similarity thresholds and unenumerated anchor and keyword lists are calibration choices that fix the working dataset size. The objective-side axiom (P.1203-based random forest from the authors' prior work, R-squared = 0.968 in [9]) is imported without re-validation here, and no objective-subjective comparison is actually computed in this preprint.

free parameters (9)
  • SHORT_TEXT_SIM_THRESHOLD = 0.40
    Hand-set cosine-similarity cutoff for short comments (fewer than 5 words) in the semantic filtering layer; no sensitivity analysis; part of the rule that selects the 33,770-comment working set.
  • SIM_THRESHOLD = 0.28
    Hand-set cosine-similarity cutoff for longer comments; with the short-text threshold, it decides which comments survive as QoE-relevant; not varied or validated.
  • MIN_WORDS = 5
    Word-count boundary switching between the 0.40 and 0.28 thresholds in the filtering pipeline.
  • anchor embedding set = not enumerated
    Predefined anchor phrases used for cosine-similarity relevance scoring; the anchor texts are not listed in the paper, making this a hidden calibration choice.
  • problem/context/meme keyword lists = not enumerated
    Curated keyword groups gate QoE relevance; examples are given but full lists are unpublished, and requiring a problem keyword biases the resulting score distribution.
  • outage threshold = -0.4 delta MOS
    Configurable delta-MOS alert threshold used to declare potential outages; treated as detection criterion without distributional or statistical justification.
  • 5-minute aggregation window = 300 seconds
    Temporal bin size for aggregation; no sensitivity analysis, and a 90-minute injected outage is far longer than the window, making the threshold crossing easy.
  • 3-second timestamp spacing = 3 seconds
    Synthetic timestamp spacing chosen to approximate a high-traffic stream; with 33,770 comments it spans about 28 hours of simulated time.
  • random seed for ISP assignment = fixed seed, value not reported
    Seed for uniform random ISP assignment; the value is not reported, so the synthetic operator labels cannot be exactly reproduced.
assumptions (6)
  • domain assumption LLM-assigned scalar MOS (1-5) is a valid measure of user-perceived quality
    Invoked in Section 4.4; supported only by citations to speech-MOS and general subjective-extraction studies [36,37]; no validation against human ratings or ITU models in the live-stream comment domain.
  • domain assumption Filtered comments are a representative sample of QoE-relevant user experience
    The filter in Section 4.3 requires a problem keyword plus a context keyword plus embedding similarity, so the retained 33,770 comments are a biased sample; the paper treats them as the population of QoE feedback.
  • domain assumption Equal user bases across ISPs and constant comment volume over time
    Stated in Section 6.1; makes windowed per-ISP MOS and global MOS directly comparable without normalization; the paper acknowledges the simplification.
  • ad hoc to paper Random ISP assignment models operator-level attribution
    Sections 4.5 and 5.3 assign ISPs uniformly at random; the paper calls this a simulation, but the operator-level analysis and the outage experiment inherit this label noise, so per-ISP signal is random plus injected.
  • ad hoc to paper Overwriting MOS to 1.0 over a time window emulates a real outage
    Eq. 4 in Section 6.4 directly sets the outcome variable; no causal link from network conditions to comments, or from comments to MOS, is simulated, so the experiment cannot test sensing.
  • domain assumption The P.1203-based random forest from [9] yields valid objective MOS
    Imported wholesale from the authors' prior work (R-squared = 0.968 on 22,000 emulated samples per [9]); not re-derived here, and no objective versus subjective comparison is computed in this preprint.
invented entities (1)
  • Delta MOS (delta MOS metric)
    purpose: Per-ISP deviation of windowed average comment MOS from the platform-wide average (Eq. 3), intended as an outage and degradation flag.
    A centered mean difference of the aggregated scores; it carries no falsifiable handle outside the paper's own synthetic data, and its crossing of the -0.4 threshold in the outage test is guaranteed by the injection in Eq. 4.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bridging Subjective and Objective QoE: Operator-Level Aggregation Using LLM-Based Comment Analysis and Network MOS Comparison." pith.science (2026). https://pith.science/paper/MMOR2QQI

@misc{pith2026250600924,
  author       = {Pith},
  title        = {Pith review of: Bridging Subjective and Objective QoE: Operator-Level Aggregation Using LLM-Based Comment Analysis and Network MOS Comparison},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MMOR2QQI}},
  note         = {Machine review of arXiv:2506.00924}
}
read the original abstract

This paper introduces a dual-layer framework for network operator-side quality of experience (QoE) assessment that integrates both objective network modeling and subjective user perception extracted from live-streaming platforms. On the objective side, we develop a machine learning model trained on mean opinion scores (MOS) computed via the ITU-T P.1203 reference implementation, allowing accurate prediction of user-perceived video quality using only network parameters such as packet loss, delay, jitter, and throughput without reliance on video content or client-side instrumentation. On the subjective side, we present a semantic filtering and scoring pipeline that processes user comments from live streams to extract performance-related feedback. A large language model is used to assign scalar MOS scores to filtered comments in a deterministic and reproducible manner. To support scalable and interpretable analysis, we construct a labeled dataset of 47,894 live-stream comments, of which about 34,000 are identified as QoE-relevant through multi-layer semantic filtering. Each comment is enriched with simulated Internet Service Provider attribution and temporally aligned using synthetic timestamps in 5-min intervals. The resulting dataset enables operator-level aggregation and time-series analysis of user-perceived quality. A delta MOS metric is proposed to measure each Internet service provider's deviation from platform-wide sentiment, allowing detection of localized degradations even in the absence of direct network telemetry. A controlled outage simulation confirms the framework's effectiveness in identifying service disruptions through comment-based trends alone. The system provides each operator with its own subjective MOS and the global platform average per interval, enabling real-time interpretation of performance deviations and comparison with objective network-based QoE estimates.

Figures

Figures reproduced from arXiv: 2506.00924 by the authors.

Figure 1
Figure 1. Semantic filtering and LLM-based MOS estimation applied to real-time live stream comments. Each entry [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Three-layer QoE model illustrating how ob [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Semantic filtering pipeline for identifying [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Subjective MOS estimation using a LLM. After [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Distribution of MOS derived from LLM evalu [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: Average MOS over time, aggregated in 5-minute windows and grouped by simulated ISPs (ISP1–ISP3). [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: Temporal analysis of ∆ MOS, calculated as the difference between each ISP’s average MOS and the global [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 8
Figure 8. Figure 8: This global reference curve, computed by averag [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]
Figure 8
Figure 8. Figure 8: Platform-wide MOS over time, averaged across all ISPs in 5-minute windows. The solid line represents [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]
Figure 9
Figure 9. Figure 9: ∆MOS time series for ISP3 in 5-minute intervals, highlighting quality deviations from the platform average. [PITH_FULL_IMAGE:figures/full_fig_p016_9.png]
Figure 10
Figure 10. Figure 10: ∆MOS time series for ISP3 with a simulated outage injected between 15:00 and 16:30 on January 1, 2024. [PITH_FULL_IMAGE:figures/full_fig_p017_10.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

42 extracted references · 40 canonical work pages

  1. [1]

    Enhancing quality of experience in telecommunica- tion networks: A review of frameworks and machine learning algorithms,

    P. H. S. Panahi, A. H. Jalilvand, and A. Diyanat, “Enhancing quality of experience in telecommunica- tion networks: A review of frameworks and machine learning algorithms,” 2024

    work page 2024

  2. [2]

    Survey on quality of experience evalua- tion for cloud-based interactive applications,

    J. Arellano-Uson, E. Maga˜ na, D. Morato, and M. Izal, “Survey on quality of experience evalua- tion for cloud-based interactive applications,” Ap- plied Sciences, vol. 14, no. 5, p. 1987, 2024. 17

    work page 1987

  3. [3]

    Investigation into user-centric qoe and network-centric parameters for youtube service on mobile networks,

    P. Anchuen and P. Uthansakul, “Investigation into user-centric qoe and network-centric parameters for youtube service on mobile networks,” in Proceedings of the 7th International Conference on Communications and Broadband Networking , ser. ICCBN ’19. New York, NY, USA: Association for Computing Machinery, 2019, p. 28–32. [Online]. Available: https://doi.o...

    work page arXiv 2019

  4. [4]

    Qoe in multimedia domain: a user-centric quality assess- ment,

    Y. Gao, X. Wei, L. Zhou, and M. Guizani, “Qoe in multimedia domain: a user-centric quality assess- ment,” Int. J. Multim. Intell. Secur. , vol. 3, pp. 162– 186, 2018

    work page 2018

  5. [5]

    Acqua: A user friendly platform for lightweight net- work monitoring and qoe forecasting,

    O. Belmoukadam, T. Spetebroot, and C. Barakat, “Acqua: A user friendly platform for lightweight net- work monitoring and qoe forecasting,” 2019 22nd Conference on Innovation in Clouds, Internet and Networks and Workshops (ICIN) , pp. 88–93, 2019

    work page 2019

  6. [6]

    Enhanc- ing qoe assessment in fwa: Leveraging network kpis and user feedback analysis,

    H. Gokcesu, O. Ercetin, and G. Kalem, “Enhanc- ing qoe assessment in fwa: Leveraging network kpis and user feedback analysis,” 2023 IEEE 28th Inter- national Workshop on Computer Aided Modeling and Design of Communication Links and Networks (CA- MAD), pp. 233–239, 2023

    work page 2023

  7. [7]

    Ucwe: A user-centric approach for web quality of experience measurement,

    X. Zheng, Y. Jin, X. Shi, Y. Song, and X. Zhao, “Ucwe: A user-centric approach for web quality of experience measurement,” 2019 IEEE Intl Conf on Parallel & Distributed Processing with Applications, Big Data & Cloud Computing, Sustainable Comput- ing & Communications, Social Computing & Net- working (ISPA/BDCloud/SocialCom/SustainCom) , pp. 928–935, 2019

    work page 2019

  8. [8]

    On user-centric modular qoe pre- diction for voip based on machine-learning algo- rithms,

    P. Charonyktakis, M. Plakia, I. Tsamardinos, and M. Papadopouli, “On user-centric modular qoe pre- diction for voip based on machine-learning algo- rithms,” IEEE Transactions on Mobile Computing , vol. 15, no. 6, pp. 1443–1456, 2016

    work page 2016

Show all 42 references
  1. [9]

    An efficient network-based qoe assess- ment framework for multimedia networks using a machine learning approach,

    P. H. S. Panahi, A. Hossein Jalilvand, and A. Diyanat, “An efficient network-based qoe assess- ment framework for multimedia networks using a machine learning approach,” IEEE Open Journal of the Communications Society , vol. 6, pp. 1653–1669, 2025

    2025

  2. [10]

    Quality of experience estimation of en- crypted video streaming by using machine learning methods,

    I. Orsolic, “Quality of experience estimation of en- crypted video streaming by using machine learning methods,” Ph.D. dissertation, 10 2020

    2020

  3. [11]

    A survey on multimedia services qoe assessment and machine learning-based prediction,

    G. Kougioumtzidis, V. Poulkov, Z. D. Zaharis, and P. I. Lazaridis, “A survey on multimedia services qoe assessment and machine learning-based prediction,” IEEE Access, vol. 10, pp. 19 507–19 538, 2022

    2022

  4. [12]

    Speech and multimedia Transmission Quality (STQ); Quality of Experience; A Monitoring Archi- tecture,

    European Telecommunications Standards Institute, “Speech and multimedia Transmission Quality (STQ); Quality of Experience; A Monitoring Archi- tecture,” ETSI, Tech. Rep. TS 103 294 V1.1.1, 2015

    2015

  5. [13]

    Survey and challenges of qoe management issues in wireless net- works,

    S. Barakovi´ c, L. Skorin-Kapov et al. , “Survey and challenges of qoe management issues in wireless net- works,” Journal of Computer Networks and Commu- nications, vol. 2013, 2013

    2013

  6. [14]

    Qoe-aware analysis and management of multimedia services in 5g and be- yond heterogeneous networks,

    M. T. Sultan and H. El Sayed, “Qoe-aware analysis and management of multimedia services in 5g and be- yond heterogeneous networks,” IEEE Access, 2023

    2023

  7. [15]

    Qoe modeling for http adaptive video streaming–a survey and open challenges,

    N. Barman and M. G. Martini, “Qoe modeling for http adaptive video streaming–a survey and open challenges,” Ieee Access, vol. 7, pp. 30 831–30 859, 2019

    2019

  8. [16]

    Cache-enabled adaptive video streaming: A qoe-based evaluation study,

    E. Liotou, D. Xenakis, V. Georgara, G. Kourounio- tis, and L. Merakos, “Cache-enabled adaptive video streaming: A qoe-based evaluation study,” Future Internet, vol. 15, no. 7, p. 221, 2023

    2023

  9. [17]

    Qoe management of multimedia streaming services in future networks: A tutorial and survey,

    A. A. Barakabitze, N. Barman, A. Ahmad, S. Zad- tootaghaj, L. Sun, M. G. Martini, and L. Atzori, “Qoe management of multimedia streaming services in future networks: A tutorial and survey,” IEEE Communications Surveys & Tutorials, vol. 22, no. 1, pp. 526–565, 2019

    2019

  10. [18]

    Machine learning based quality of experience (qoe) prediction approach in enterprise multimedia networks,

    H. O. Hamidou, J. P. Kouraogo, O. Sie, and D. Tap- soba, “Machine learning based quality of experience (qoe) prediction approach in enterprise multimedia networks,” in Proceedings of the 5th edition of the Computer Science Research Days, JRI 2022, 24-26 November 2022, Ouagadou...

    2022

  11. [19]

    Llm4band: Enhancing reinforcement learning with large language models for accurate bandwidth estimation,

    Z. Wang, R. Lu, Z. Zhang, C. Westphal, D. He, and J. Jiang, “Llm4band: Enhancing reinforcement learning with large language models for accurate bandwidth estimation,” in Proceedings of the 35th Workshop on Network and Operating System Support for Digital Audio and Video , ser....

    2025

  12. [20]

    REAL-TIME MON- ITORING AND ENHANCEMENT OF MULTI- USER COMMUNICATIONS WITH AI AGENTS,

    A. Rezaee and S. Gundavelli, “REAL-TIME MON- ITORING AND ENHANCEMENT OF MULTI- USER COMMUNICATIONS WITH AI AGENTS,” https://www.tdcommons.org/dpubs series/7983, April 2025, technical Disclosure Commons, Defen- sive Publications Series

    2025

  13. [21]

    Quality-of-service aware llm serving

    S. Jha and R. Chuppala, “Quality-of-service aware llm serving. ”

  14. [22]

    Large language model aided qos prediction for service recommendation,

    H. Liu, Z. Zhang, H. Li, Q. Wu, and Y. Zhang, “Large language model aided qos prediction for service recommendation,” arXiv preprint arXiv:2408.02223, 2024

    2024 arXiv

  15. [23]

    Quality measurement of screen images via foreground perception and background suppression,

    M. Wang, Y. Huang, J. Lin, W. Xie, G. Yue, S. Wang, and L. Li, “Quality measurement of screen images via foreground perception and background suppression,” IEEE Transactions on Instrumenta- tion and Measurement, vol. 70, pp. 1–11, 2021. 18

    2021

  16. [24]

    Fine- grained region adaptive loop filter for super-block video coding,

    M. Wang, J. Lin, J. Zhang, and W. Xie, “Fine- grained region adaptive loop filter for super-block video coding,” IEEE Access, vol. 8, pp. 445–454, 2019

    2019

  17. [25]

    Quality of experience pre- diction model for progressive downloading over mo- bile broadcast networks,

    Z. Yetgin and Z. G¨ o¸ cer, “Quality of experience pre- diction model for progressive downloading over mo- bile broadcast networks,” Telecommunication Sys- tems, vol. 58, pp. 55–66, 01 2014

    2014

  18. [26]

    Mmlscu: A dataset for multi-modal multi-domain live streaming comment understanding,

    Z. Meng, Q. Gao, D. Guo, Y. Li, B. Li, H. Fei, S. Wu, F. Li, C. Teng, and D. Ji, “Mmlscu: A dataset for multi-modal multi-domain live streaming comment understanding,” in Proceedings of the ACM Web Conference 2024 , ser. WWW ’24. New York, NY, USA: Association for Computing Ma...

    2024

  19. [27]

    Time- sync comments analyzation for understanding sub- scribers to live streaming services,

    F.-L. Huang, G.-Q. Xie, and Z.-W. Chen, “Time- sync comments analyzation for understanding sub- scribers to live streaming services,” 2020

    2020

  20. [28]

    Enhanc- ing multimodal affective analysis with learned live comment features,

    Z. Deng, A. Ananthram, and K. McKeown, “Enhanc- ing multimodal affective analysis with learned live comment features,” in AAAI Conference on Artifi- cial Intelligence, 2024

    2024

  21. [29]

    Of- fensive language detection on video live streaming chat,

    Z. Gao, S. Yada, S. Wakamiya, and E. Aramaki, “Of- fensive language detection on video live streaming chat,” in International Conference on Computational Linguistics, 2020

    2020

  22. [30]

    Twitchchat: A dataset for exploring livestream chat,

    C. Ringer, M. A. Nicolaou, and J. Walker, “Twitchchat: A dataset for exploring livestream chat,” in Proceedings of the AAAI Conference on Ar- tificial Intelligence and Interactive Digital Entertain- ment, vol. 16, no. 1, 2020, pp. 259–265

    2020

  23. [31]

    Twitch.tv Chat Log Data,

    J. Kim, “Twitch.tv Chat Log Data,” 2019. [Online]. Available: https://doi.org/10.7910/DVN/VE0IVQ

    2019 doi

  24. [32]

    Seman- tic collaborative filtering recommender system using cnns,

    A. Y. Zaremarjal and D. Yiltas-Kaplan, “Seman- tic collaborative filtering recommender system using cnns,” in 2021 8th International Conference on Elec- trical and Electronics Engineering (ICEEE) , 2021, pp. 254–258

    2021

  25. [33]

    Conex based qoe feedback to enhance qoe,

    M. Shirazipour, G. Charlot, G. Lefebvre, S. Kr- ishnan, and S. Pierre, “Conex based qoe feedback to enhance qoe,” in Proceedings of the 2012 ACM Workshop on Capacity Sharing , ser. CSWS ’12. New York, NY, USA: Association for Computing Machinery, 2012, p. 27–32. [Online]. Avai...

    2012

  26. [34]

    Document-level text simplification,

    D. K. Kontoe, “Document-level text simplification,” 2023

    2023

  27. [35]

    Performance of 4 pre-trained sentence transformer models in the semantic query of a systematic review dataset on peri-implantitis,

    C. Galli, N. Donos, and E. Calciolari, “Performance of 4 pre-trained sentence transformer models in the semantic query of a systematic review dataset on peri-implantitis,” Information, vol. 15, no. 2,

  28. [36]

    Enabling auditory large language models for automatic speech quality evaluation,

    S. Wang, W. Yu, Y. Yang, C. Tang, Y. Li, J. Zhuang, X. Chen, X. Tian, J. Zhang, G. Sun et al., “Enabling auditory large language models for automatic speech quality evaluation,” in ICASSP 2025-2025 IEEE In- ternational Conference on Acoustics, Speech and Sig- nal Processing (I...

    2025

  29. [37]

    Extraction of subjective information from large language models,

    A. Kobayashi and S. Yamaguchi, “Extraction of subjective information from large language models,” in 2024 IEEE 48th Annual Computers, Software, and Applications Conference (COMPSAC), 2024, pp. 1612–1617

    2024

  30. [38]

    The past, present and better future of feedback learning in large language models for subjective hu- man preferences and values,

    H. Kirk, A. Bean, B. Vidgen, P. Rottger, and S. Hale, “The past, present and better future of feedback learning in large language models for subjective hu- man preferences and values,” 01 2023, pp. 2409–2430

    2023

  31. [39]

    A survey of un- certainty estimation methods on large language mod- els,

    Z. Xia, J. Xu, Y. Zhang, and H. Liu, “A survey of un- certainty estimation methods on large language mod- els,” arXiv preprint arXiv:2503.00172 , 2025

    2025 arXiv

  32. [40]

    Language models are few-shot learners,

    T. B. Brown, B. Mann, N. Ryder et al. , “Language models are few-shot learners,” Advances in neural information processing systems , vol. 33, pp. 1877– 1901, 2020

    1901

  33. [41]

    Gpt-4o overview and capabilities,

    OpenAI, “Gpt-4o overview and capabilities,” https: //openai.com/index/hello-gpt-4o, 2024, accessed: May 2025. 19

    2024

  34. [2024]

    Available: https://www.mdpi.com/ 2078-2489/15/2/68

    [Online]. Available: https://www.mdpi.com/ 2078-2489/15/2/68

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.