Retrieval-Augmented Explainable Fake Review Detection with Evidence-Grounded Indicators
Keywords:
fake review detection, explainable AI, hybrid retrievalAbstract
Fake review detectors must be accurate and able to justify their decisions to moderators, sellers and regulators. Knowledge-enhanced detectors built on large language models (LLMs) have shown that curated fraud knowledge and step-by-step rationales improve both accuracy and interpretability, but they rely on GPU fine-tuning and generative inference. This paper presents GRADE (Grounded Retrieval-Augmented Detection with Explanations), a lightweight framework in which every decision is grounded in three verifiable sources of evidence: labelled precedent reviews retrieved by a hybrid sparse-dense retriever over pretrained language model embeddings, a knowledge base of 25 executable fraud indicators (linguistic, sentiment and behavioural) calibrated on the training data, and a sparse lexical model that exposes the n-grams driving the decision. An additive stacker combines the three channels, so the log-odds of each verdict decompose exactly into channel contributions and the explanation is faithful by construction. On two public benchmarks, human-written deceptive hotel reviews and machine-generated Amazon reviews, GRADE reaches 90.0% and 95.2% accuracy (AUC-ROC 96.7% and 99.0%), matching or exceeding the strongest non-retrieval baselines; it is up to 5.3 F1 points more accurate when only a few hundred labelled reviews are available, transfers across product categories with up to 8.7 F1 points more than text-only models, and runs at 54–150 ms per review on a single CPU core. Perturbation tests show that the evidence cited by GRADE, obtained at no extra cost, approaches the faithfulness of a LIME surrogate that needs 400 additional model evaluations per review and is three to nine times more faithful than random evidence at the same budget.
Downloads
References
[1] M. Luca and G. Zervas, “Fake it till you make it: Reputation, competition, and Yelp review fraud,” Management Science, vol. 62, no. 12, pp. 3412–3427, 2016.
[2] Y. Wu, E. W. T. Ngai, P. Wu, and C. Wu, “Fake online reviews: Literature review, synthesis, and directions for future research,” Decision Support Systems, vol. 132, p. 113280, 2020.
[3] R. Mohawesh, S. Xu, S. N. Tran, R. Ollington, M. Springer, Y. Jararweh, and S. Maqsood, “Fake reviews detection: A survey,” IEEE Access, vol. 9, pp. 65771–65802, 2021.
[4] M. Ott, Y. Choi, C. Cardie, and J. T. Hancock, “Finding deceptive opinion spam by any stretch of the imagination,” in Proc. 49th Annu. Meeting Assoc. Comput. Linguistics (ACL), 2011, pp. 309–319.
[5] J. Wang, Y. Ge, J. Liu, and B. Jiang, “Early warning of liquidity stress in U.S. banks with seasonally aligned frailty hazards and delayed-feedback conformal alarms,” Journal of Trends in Finance and Economics, vol. 1, no. 1, pp. 85–100, 2024.
[6] S. Rayana and L. Akoglu, “Collective opinion spam detection: Bridging review networks and metadata,” in Proc. 21st ACM SIGKDD Int. Conf. Knowl. Discovery Data Mining (KDD), 2015, pp. 985–994.
[7] Y. Ren and D. Ji, “Neural networks for deceptive opinion spam detection: An empirical study,” Information Sciences, vol. 385–386, pp. 213–224, 2017.
[8] J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “BERT: Pre-training of deep bidirectional transformers for language understanding,” in Proc. NAACL-HLT, 2019, pp. 4171–4186.
[9] Y. Dou, Z. Liu, L. Sun, Y. Deng, H. Peng, and P. S. Yu, “Enhancing graph neural network-based fraud detectors against camouflaged fraudsters,” in Proc. 29th ACM Int. Conf. Inf. Knowl. Manage. (CIKM), 2020, pp. 315–324.
[10] E. J. Hu, Y. Shen, P. Wallis, Z. Allen-Zhu, Y. Li, S. Wang, L. Wang, and W. Chen, “LoRA: Low-rank adaptation of large language models,” in Proc. Int. Conf. Learn. Representations (ICLR), 2022.
[11] J. Wei, X. Wang, D. Schuurmans, M. Bosma, B. Ichter, F. Xia, E. Chi, Q. V. Le, and D. Zhou, “Chain-of-thought prompting elicits reasoning in large language models,” in Advances in Neural Information Processing Systems, vol. 35, 2022, pp. 24824–24837.
[12] S. Jain and B. C. Wallace, “Attention is not explanation,” in Proc. NAACL-HLT, 2019, pp. 3543–3556.
[13] A. Jacovi and Y. Goldberg, “Towards faithfully interpretable NLP systems: How should we define and evaluate faithfulness?” in Proc. 58th Annu. Meeting Assoc. Comput. Linguistics (ACL), 2020, pp. 4198–4205.
[14] P. Lewis, E. Perez, A. Piktus, F. Petroni, V. Karpukhin, N. Goyal, H. Küttler, M. Lewis, W.-t. Yih, T. Rocktäschel, S. Riedel, and D. Kiela, “Retrieval-augmented generation for knowledge-intensive NLP tasks,” in Advances in Neural Information Processing Systems, vol. 33, 2020, pp. 9459–9474.
[15] O. Ram, Y. Levine, I. Dalmedigos, D. Muhlgay, A. Shashua, K. Leyton-Brown, and Y. Shoham, “In-context retrieval-augmented language models,” Transactions of the Association for Computational Linguistics, vol. 11, pp. 1316–1331, 2023.
[16] D. Teng, “TRIAD: Deadline- and reliability-aware scheduling for post-earthquake structural assessment workflows,” World Journal of Engineering Research, vol. 3, no. 6, pp. 55–66, 2025.
[17] J. Liu, D. Shen, Y. Zhang, B. Dolan, L. Carin, and W. Chen, “What makes good in-context examples for GPT-3?” in Proc. Deep Learning Inside Out (DeeLIO 2022): 3rd Workshop Knowl. Extraction Integration Deep Learn. Archit., 2022, pp. 100–114.
[18] J. Salminen, C. Kandpal, A. M. Kamel, S.-g. Jung, and B. J. Jansen, “Creating and detecting fake reviews of online products,” Journal of Retailing and Consumer Services, vol. 64, p. 102771, 2022.
[19] J. Chen, J. Liu, Y. Liang, and M. Zhou, “KE-MLLM: A knowledge-enhanced multi-sensor learning framework for explainable fake review detection,” Applied Sciences, vol. 16, no. 6, p. 2909, 2026.
[20] K. Guu, K. Lee, Z. Tung, P. Pasupat, and M.-W. Chang, “REALM: Retrieval-augmented language model pre-training,” in Proc. 37th Int. Conf. Mach. Learn. (ICML), PMLR vol. 119, 2020, pp. 3929–3938.
[21] A. B. Arrieta, N. Díaz-Rodríguez, J. Del Ser, A. Bennetot, S. Tabik, A. Barbado, S. García, S. Gil-López, D. Molina, R. Benjamins, R. Chatila, and F. Herrera, “Explainable artificial intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI,” Information Fusion, vol. 58, pp. 82–115, 2020.
[22] M. T. Ribeiro, S. Singh, and C. Guestrin, “‘Why should I trust you?’ Explaining the predictions of any classifier,” in Proc. 22nd ACM SIGKDD Int. Conf. Knowl. Discovery Data Mining (KDD), 2016, pp. 1135–1144.
[23] J. DeYoung, S. Jain, N. F. Rajani, E. Lehman, C. Xiong, R. Socher, and B. C. Wallace, “ERASER: A benchmark to evaluate rationalized NLP models,” in Proc. 58th Annu. Meeting Assoc. Comput. Linguistics (ACL), 2020, pp. 4443–4458.
[24] W. Wang, F. Wei, L. Dong, H. Bao, N. Yang, and M. Zhou, “MiniLM: Deep self-attention distillation for task-agnostic compression of pre-trained transformers,” in Advances in Neural Information Processing Systems, vol. 33, 2020, pp. 5776–5788.
[25] N. Reimers and I. Gurevych, “Sentence-BERT: Sentence embeddings using Siamese BERT-networks,” in Proc. Conf. Empirical Methods Natural Language Processing and 9th Int. Joint Conf. Natural Language Processing (EMNLP-IJCNLP), 2019, pp. 3982–3992.
[26] S. Robertson and H. Zaragoza, “The probabilistic relevance framework: BM25 and beyond,” Foundations and Trends in Information Retrieval, vol. 3, no. 4, pp. 333–389, 2009.
[27] G. V. Cormack, C. L. A. Clarke, and S. Büttcher, “Reciprocal rank fusion outperforms Condorcet and individual rank learning methods,” in Proc. 32nd Int. ACM SIGIR Conf. Res. Develop. Inf. Retrieval, 2009, pp. 758–759.
[28] C. J. Hutto and E. Gilbert, “VADER: A parsimonious rule-based model for sentiment analysis of social media text,” in Proc. 8th Int. AAAI Conf. Weblogs Social Media (ICWSM), 2014, pp. 216–225.
[29] D. H. Wolpert, “Stacked generalization,” Neural Networks, vol. 5, no. 2, pp. 241–259, 1992.
[30] M. Ott, C. Cardie, and J. T. Hancock, “Negative deceptive opinion spam,” in Proc. Conf. North Amer. Chapter Assoc. Comput. Linguistics: Human Language Technologies (NAACL-HLT), 2013, pp. 497–501.
[31] F. Zhou, “Moment-constrained side-force allocation for wind-assisted merchant ships: Hydrodynamic performance of retractable anti-leeway fins and flap rudders,” Innovation and Technology Studies, vol. 3, no. 1, pp. 27–41, 2026.
[32] M. Rhee et al., “Enhancing Web Search Agents With Self-Play Contrastive Fine-Tuning,” IEEE Access, 2026, doi: 10.1109/ACCESS.2026.3717594.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Riley Lin, Liam Cooper, Thomas Yoon
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
This article is licensed under the Creative Commons Attribution 4.0 International License. This license permits use, sharing, and adaptation of the article for non-commercial purposes, provided the original authors and source are properly credited.
Alternatively, you may choose other Creative Commons licenses based on your preferences:
-
CC Attribution-NonCommercial-NoDerivatives 4.0: This license allows others to download the works and share them with others as long as they credit the author, but they can’t change them in any way or use them commercially.
-
CC Attribution-NonCommercial 4.0: This license allows others to remix, tweak, and build upon the work non-commercially, as long as they credit the author, but they cannot use the material for commercial purposes.
-
CC Attribution-NonCommercial-ShareAlike 4.0: This allows others to remix, tweak, and build upon the work non-commercially, as long as they credit the author and license their new creations under identical terms.
These terms ensure that your article is freely accessible and usable, while respecting the rights of the original authors.
