pith. sign in

arxiv: 1907.05028 · v1 · pith:QEMYCJIVnew · submitted 2019-07-11 · 💻 cs.SI · cs.AI

Evidential positive opinion influence measures for viral marketing

Siwar Jendoubi (LARODEC) , Arnaud Martin (DRUID) This is my paper

Pith reviewed 2026-05-24 23:04 UTC · model grok-4.3

classification 💻 cs.SI cs.AI
keywords viral marketinginfluence maximizationevidential theoryopinion influencesocial networkspositive opinionsTwitter dataset
0
0 comments X

The pith

Evidential theory supplies six influence measures to detect positive-opinion influencers across three viral marketing scenarios.

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

The paper builds an influence maximization model that treats opinions as evidence masses to locate influencers who can trigger product adoption. It isolates three concrete scenarios: influencers holding positive opinions, those whose positive opinions reinforce other positive opinions, and those whose positive opinions convert negative opinions. For each scenario the authors define two dedicated influence measures and apply a standard maximization procedure to select the top influencers. Experiments on a synthetic network and a Twitter crawl show that the resulting sets differ by scenario and produce measurable spread under each measure.

Core claim

By representing user opinions inside an evidential framework, the authors define six influence measures (two per scenario) that quantify how positive opinions propagate or flip negative ones; these measures are then inserted into an influence maximization routine that returns a distinct seed set for each of the three scenarios, and the procedure is shown to run on both generated graphs and a real Twitter dataset.

What carries the argument

Six evidential influence measures (two per scenario) that assign mass to positive, negative, and uncertain opinions and compute a scalar influence score from those masses.

If this is right

  • Each of the three scenarios yields its own ranked list of influencers rather than a single universal ranking.
  • The maximization step can be rerun independently for each pair of measures to produce scenario-specific seed sets.
  • Performance differences among the six measures can be compared directly on the same generated and Twitter graphs.
  • The model distinguishes reinforcement of positive opinions from conversion of negative opinions as separate influence tasks.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Campaign planners could run the three scenarios in parallel and allocate budget according to which opinion dynamic matters most for their product.
  • The measures might be recomputed after each wave of observed adoptions to update the influencer ranking in an online setting.
  • If the evidential masses are estimated from richer text or sentiment data, the same six-measure structure could be reused without redesigning the maximization step.

Load-bearing premise

Evidential theory can be applied directly to map observed opinions onto evidence masses inside real social networks without further validation of that mapping.

What would settle it

On the Twitter dataset, the influencer sets returned by the six measures produce adoption cascades whose size is statistically indistinguishable from those produced by degree or random baselines under the same diffusion model.

Figures

Figures reproduced from arXiv: 1907.05028 by Arnaud Martin (DRUID), Siwar Jendoubi (LARODEC).

Figure 1
Figure 1. Figure 1: Data distributions [15] 21 [PITH_FULL_IMAGE:figures/full_fig_p021_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Comparison between the opinion based scenarios, the second influence model and the OC model [PITH_FULL_IMAGE:figures/full_fig_p030_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Accuracy variation while varying the minimum influence value [PITH_FULL_IMAGE:figures/full_fig_p034_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Accuracy of detected positive influencers while varying the minimum positive opinion [PITH_FULL_IMAGE:figures/full_fig_p035_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Accuracy of detected positive influencers influencing positive users while varying [PITH_FULL_IMAGE:figures/full_fig_p036_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Accuracy of detected positive influencers influencing positive and negative users [PITH_FULL_IMAGE:figures/full_fig_p037_6.png] view at source ↗
read the original abstract

The Viral Marketing is a relatively new form of marketing that exploits social networks to promote a brand, a product, etc. The idea behind it is to find a set of influencers on the network that can trigger a large cascade of propagation and adoptions. In this paper, we will introduce an evidential opinion-based influence maximization model for viral marketing. Besides, our approach tackles three opinions based scenarios for viral marketing in the real world. The first scenario concerns influencers who have a positive opinion about the product. The second scenario deals with influencers who have a positive opinion about the product and produce effects on users who also have a positive opinion. The third scenario involves influence users who have a positive opinion about the product and produce effects on the negative opinion of other users concerning the product in question. Next, we proposed six influence measures, two for each scenario. We also use an influence maximization model that the set of detected influencers for each scenario. Finally, we show the performance of the proposed model with each influence measure through some experiments conducted on a generated dataset and a real world dataset collected from Twitter.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper proposes an evidential opinion-based influence maximization model for viral marketing that handles three scenarios: (1) positive-opinion influencers, (2) positive influencers affecting positive-opinion users, and (3) positive influencers converting negative-opinion users. It defines six influence measures (two per scenario) using Dempster-Shafer evidential theory, applies an influence maximization procedure to select seed sets for each scenario, and reports experimental comparisons against baselines on one synthetic network and one Twitter dataset.

Significance. If the opinion-to-belief-mass mappings prove robust, the work would offer a principled way to incorporate uncertainty and opinion polarity into influence maximization, extending standard IM models to the three realistic marketing scenarios described. The use of both generated and real Twitter data, together with the explicit separation of the three scenarios, would be a concrete contribution to opinion-aware viral marketing.

major comments (2)
  1. [Section 3 (influence measure definitions)] The central claim that the six evidential measures correctly distinguish the three positive-opinion scenarios rests on an unvalidated mapping from observed opinions to basic belief assignments. No explicit formulas for the mass functions m({+}), m({−}), m(Θ) or justification that they capture the intended semantics (positive-to-positive effects vs. positive-to-negative conversion) appear in the methods; without this, the measures and subsequent maximization lack grounding.
  2. [Section 5] Section 5 (experiments): the reported outperformance on the Twitter dataset is not accompanied by sensitivity checks on alternative belief-mass assignments or by ablation of the evidential combination rule. If results change materially under plausible reparameterizations of the masses, the claim that the measures are superior to baselines cannot be sustained.
minor comments (2)
  1. Notation for the three scenarios and the six measures is introduced without a summary table; a single table listing scenario, measure name, and key formula would improve readability.
  2. The abstract is lengthy and repeats the scenario descriptions; condensing it would help readers quickly grasp the contribution.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback on our manuscript. We address each major comment point by point below, clarifying the current content of the paper and indicating where revisions will be incorporated to strengthen the presentation and validation.

read point-by-point responses

  1. Referee: [Section 3 (influence measure definitions)] The central claim that the six evidential measures correctly distinguish the three positive-opinion scenarios rests on an unvalidated mapping from observed opinions to basic belief assignments. No explicit formulas for the mass functions m({+}), m({−}), m(Θ) or justification that they capture the intended semantics (positive-to-positive effects vs. positive-to-negative conversion) appear in the methods; without this, the measures and subsequent maximization lack grounding.

    Authors: We thank the referee for this observation. Section 3 introduces the six measures using Dempster-Shafer theory and describes how opinion polarities are mapped to belief masses for each of the three scenarios, but we agree that the explicit formulas for m({+}), m({−}), and m(Θ) and the semantic justification linking them to positive-to-positive influence versus conversion effects are not presented with sufficient detail or tabular clarity. We will revise Section 3 to include the precise mass assignment formulas for each scenario together with a dedicated paragraph explaining how the assignments encode the intended semantics. revision: yes

  2. Referee: [Section 5] Section 5 (experiments): the reported outperformance on the Twitter dataset is not accompanied by sensitivity checks on alternative belief-mass assignments or by ablation of the evidential combination rule. If results change materially under plausible reparameterizations of the masses, the claim that the measures are superior to baselines cannot be sustained.

    Authors: The referee correctly notes the absence of these checks. The experiments in Section 5 rely on a single opinion-to-mass mapping and the standard Dempster combination rule without exploring alternatives or performing ablation. We will add a new subsection to Section 5 that reports sensitivity results under varied mass assignments (including different uncertainty levels for m(Θ)) and an ablation comparing the full evidential model against direct opinion aggregation without the combination rule, using the same Twitter dataset. revision: yes

Circularity Check

0 steps flagged

No circularity; derivation introduces new measures without reduction to fitted inputs or self-citations

full rationale

The abstract and available text describe the proposal of six new influence measures grounded in evidential theory applied to three opinion scenarios, followed by an influence maximization model and experimental validation on generated and Twitter datasets. No equations, parameter-fitting procedures, or self-citation chains are presented that would reduce any claimed prediction or result to the inputs by construction. The work positions itself as extending evidential theory to new scenarios rather than deriving outputs from prior fitted values or author-specific uniqueness theorems. This is the most common honest outcome when no load-bearing self-referential steps are visible.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review supplies no information on free parameters, axioms, or invented entities.

pith-pipeline@v0.9.0 · 5722 in / 1052 out tokens · 21096 ms · 2026-05-24T23:04:57.482397+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

34 extracted references · 34 canonical work pages

  1. [1]

    In: Proceed- ings of the Seventh conference on International Language Resources and Evaluation

    Baccianella, S., Esuli, A., Sebatiani, F.: Sentiwordnet 3.0: An enhanced lexical resource for sentiment analysis and opinion mining. In: Proceed- ings of the Seventh conference on International Language Resources and Evaluation. pp. 2200–2204 (May 2010)

    work page 2010

  2. [2]

    Knowledge and Information Systems 37(3), 555–584 (Decem- ber 2013)

    Barbieri, N., Bonchi, F., Manco, G.: Topic-aware social influence propaga- tion models. Knowledge and Information Systems 37(3), 555–584 (Decem- ber 2013)

    work page 2013

  3. [3]

    Review of General Psychology 5(4), 323–270 (2001)

    Baumeister, R., Bratslavsky, E., Finkenauer, C., Vohs, K.: Bad is stronger than good. Review of General Psychology 5(4), 323–270 (2001)

    work page 2001

  4. [4]

    Physica A: Statistical mechanics and its applications 391(4), 1777–1787 (2012)

    Chen, D., Lü, L., Shang, M.S., Zhang, Y.C., Zhou, T.: Identifying influ- ential nodes in complex networks. Physica A: Statistical mechanics and its applications 391(4), 1777–1787 (2012)

    work page 2012

  5. [5]

    Chen, W., Collins, A., Cummings, R., Ke, T., Liu, Z., Rincon, D., Sun, X., Wang, Y., Wei, W., Yuan, Y.: Influence maximization in social networks whennegativeopinionsmayemergeandpropagate.In: ProcedingsofSIAM SDM. pp. 379–390 (April 2011)

    work page 2011

  6. [6]

    Cheung, C.M., Lee, M.K.: Online consumer reviews: Does negative elec- tronicword-of-mouthhurtmore? In: Proceedingofthefourteenthamericas conference on information systems. p. 143 (August 2008)

    work page 2008

  7. [7]

    Annals of Mathematical Statistics 38, 325–339 (1967)

    Dempster, A.P.: Upper and Lower probabilities induced by a multivalued mapping. Annals of Mathematical Statistics 38, 325–339 (1967)

    work page 1967

  8. [8]

    In: Proceedings of KDD’01

    Domingos, P., Richardson, M.: Mining the network value of customers. In: Proceedings of KDD’01. pp. 57–66 (2001)

    work page 2001

  9. [9]

    Physica A 392(21), 5490–5500 (November 2013) 39

    Gao, C., Wei, D., Hu, Y., Mahadevan, S., Deng, Y.: A modified evidential methodology of identifying influential nodes in weighted networks. Physica A 392(21), 5490–5500 (November 2013) 39

    work page 2013

  10. [10]

    Goldenberg, J., Libai, B., Muller, E.: Talk of the network: A complex systemslookattheunderlyingprocessofword-of-mouth.MarketingLetters 12(3), 211–223 (August 2001)

    work page 2001

  11. [11]

    In: Proceedings of VLDB Endowment

    Goyal, A., Bonchi, F., Lakshmanan, L.V.S.: A data-based approach to social influence maximization. In: Proceedings of VLDB Endowment. pp. 73–84 (August 2012)

    work page 2012

  12. [12]

    American jour- nal of sociology pp

    Granovetter, M.: Threshold models of collective behavior. American jour- nal of sociology pp. 1420–1443 (1978)

    work page 1978

  13. [13]

    In: Belief Functions: Theory and Applica- tions, pp

    Jendoubi, S., Chebbah, M., Martin, A.: Evidential independence maxi- mization on twitter network. In: Belief Functions: Theory and Applica- tions, pp. 121–128. Springer International Publishing (2018)

    work page 2018

  14. [14]

    In: Poceeding of the 12th international FLINS conference (August 2016)

    Jendoubi, S., Martin, A., Liétard, L., Ben Hadj, H., Ben Yaghlane, B.: Maximizing positive opinion influence using an evidential approach. In: Poceeding of the 12th international FLINS conference (August 2016)

    work page 2016

  15. [15]

    Knowledge-Based Systems 121, 58–70 (2017)

    Jendoubi, S., Martin, A., Liétard, L., Ben Hadj, H., Ben Yaghlane, B.: Two evidential data based models for influence maximization in twitter. Knowledge-Based Systems 121, 58–70 (2017)

    work page 2017

  16. [16]

    In: Proceeding of IPMU

    Jendoubi, S., Martin, A., Liétard, L., Ben Yaghlane, B.: Classification of message spreading in a heterogeneous social network. In: Proceeding of IPMU. pp. 66–75 (July 2014)

    work page 2014

  17. [17]

    In: Proceeding of ECSQARU

    Jendoubi, S., Martin, A., Liétard, L., Ben Yaghlane, B., Ben Hadj, H.: Dynamic time warping distance for message propagation classification in twitter. In: Proceeding of ECSQARU. pp. 419–428 (July 2015)

    work page 2015

  18. [18]

    Jurvetson, S.: What exactly is viral marketing? Red Herring 78, 110–112 (2000)

    work page 2000

  19. [19]

    In: Proceedings of KDD’03

    Kempe, D., Kleinberg, J., Tardos, E.: Maximizing the spread of influence through a social network. In: Proceedings of KDD’03. pp. 137–146 (August 2003) 40

    work page 2003

  20. [20]

    In: Proceed- ings of KDD’07

    Leskovec, J., Krause, A., Guestrin, C., Faloutsos, C., VanBriesen, J., Glance, N.: Cost-effective outbreak detection in networks. In: Proceed- ings of KDD’07. pp. 420–429 (August 2007)

    work page 2007

  21. [21]

    PLoS ONE 9(7), e102199 (July 2014)

    Li, D., Xu, Z.M., Chakraborty, N., Gupta, A., Sycara, K., Li, S.: Polarity related influence maximization in signed social networks. PLoS ONE 9(7), e102199 (July 2014)

    work page 2014

  22. [22]

    Information Sciences 181(23), 5143–5157 (December 2011)

    Li, Y.M., Lai, C.Y., Chen, C.W.: Discovering influencers for marketing in the blogosphere. Information Sciences 181(23), 5143–5157 (December 2011)

    work page 2011

  23. [23]

    ACM Transactions on Knowledge Discovery from Data 11(3), 1–30 (mar 2017)

    Liu, Q., Xiang, B., Yuan, N.J., Chen, E., Xiong, H., Zheng, Y., Yang, Y.: An influence propagation view of PageRank. ACM Transactions on Knowledge Discovery from Data 11(3), 1–30 (mar 2017)

    work page 2017

  24. [24]

    Knowledge and Information Systems 51(1), 159–186 (April 2017)

    Moosavi, S.A., Jalali, M., Misaghian, N., Shamshirband, S., Anisi, M.H.: Communitydetectioninsocialnetworksusinguserfrequentpatternmining. Knowledge and Information Systems 51(1), 159–186 (April 2017)

    work page 2017

  25. [25]

    Narayanam, R., Nanavati, A.A.: Design of viral marketing strategies for productcross-sellthroughsocialnetworks.KnowledgeandInformationSys- tems 39(3), 609–641 (June 2014)

    work page 2014

  26. [26]

    Oxford University Press (2010)

    Newman, M.E.J.: Networks: An introduction. Oxford University Press (2010)

    work page 2010

  27. [27]

    Princeton University Press (1976)

    Shafer, G.: A mathematical theory of evidence. Princeton University Press (1976)

    work page 1976

  28. [28]

    Artificial Intelligent 66, 191–234 (1994)

    Smets, P., Kennes, R.: The Transferable Belief Model. Artificial Intelligent 66, 191–234 (1994)

    work page 1994

  29. [29]

    Psychological bulletin 1(110), 67–85 (1991) 41

    Taylor, S.E.: Asymmetrical effects of positive and negative events: the mobilization-minimization hypothesis. Psychological bulletin 1(110), 67–85 (1991) 41

    work page 1991

  30. [30]

    Physica A 392(10), 2564–2575 (Mai 2013)

    Wei, D., Deng, X., Zhang, X., Deng, Y., Mahadeven, S.: Identifying in- fluential nodes in weighted networks based on evidence theory. Physica A 392(10), 2564–2575 (Mai 2013)

    work page 2013

  31. [31]

    In: Proceedings of the Twenty-Third International Joint Conference on Artificial Intelligence

    Xiang, B., Liu, Q., Chen, E., Xiong, H., Zheng, Y., Yang, Y.: Pagerank with priors: An influence propagation perspective. In: Proceedings of the Twenty-Third International Joint Conference on Artificial Intelligence. pp. 2740–2746 (2013)

    work page 2013

  32. [32]

    IEEE Transactions on Knowledge and Data Engineering 30(5), 810–823 (may 2018)

    Yang, J., Liu, C., Teng, M., Chen, J., Xiong, H.: A unified view of so- cial and temporal modeling for b2b marketing campaign recommendation. IEEE Transactions on Knowledge and Data Engineering 30(5), 810–823 (may 2018)

    work page 2018

  33. [33]

    In: Proceedings of ICDCS

    Zhang, H., Dinh, T.N., Thai, M.T.: Maximizing the spread of positive influence in online social networks. In: Proceedings of ICDCS. pp. 317–326 (July 2013)

    work page 2013

  34. [34]

    Knowledge-Based Sys- tems 74, 69–88 (2015) 42 Siwar Jendoubi is a researcher at LARODEC Laboratory

    Zhou, K., Martin, A., Pan, Q., Liu, Z.g.: Median evidential c-means algo- rithm and its application to community detection. Knowledge-Based Sys- tems 74, 69–88 (2015) 42 Siwar Jendoubi is a researcher at LARODEC Laboratory. She received a PhD degree (2016) in computer science and a Master degree (2012) in business intelligence. Dr. Siwar Jendoubi joined t...

    work page 2015