The reviewed record of science sign in
Pith

arxiv: 2606.28002 · v1 · pith:JUDHW35M · submitted 2026-06-26 · cs.CL · cs.AI· eess.AS

Dialogue to Detection: A Multimodal Hybrid NLP Pipeline for Insurance Fraud Detection

Muhammad Shakeel Akram , Amal Htait , Abdul Hamid Sadka , Emma Meisingseth , Karishma Jaitly This is my paper

Reviewed by Pith2026-06-29 04:25 UTCgrok-4.3pith:JUDHW35Mopen to challenge →

classification cs.CL cs.AIeess.AS
keywords insurance fraud detectionmultimodal NLPsynthetic dataFNOLdialogue generationspeaker embeddingsrule-based scoringfraud pipeline
0
0 comments X

The pith

A synthetic multimodal framework generates FNOL dialogues and audio to detect insurance fraud with a hybrid NLP pipeline.

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

The paper presents a synthetic multimodal framework that replicates first notice of loss conditions by creating agent-customer dialogue transcripts and two-speaker audio files. It processes these through automatic speech recognition and diarisation, then applies named entity recognition, regex feature extraction, LLM-RAG retrieval, and speaker embeddings inside a rule-based risk score. The score is designed to identify narrative reuse, structural inconsistencies, and repeated voices across cases while managing sensitivity and false positives. This matters because existing fraud detection relies on limited private text datasets, and multimodal signals could improve early detection without those constraints. Component evaluations indicate the modules operate stably and show transfer potential to real data.

Core claim

We introduce a synthetic multimodal framework that replicates FNOL conditions. It generates agent-customer dialogue transcripts and two-speaker audios, performs ASR and diarisation. Downstream modules combine NER, regex-based feature extraction, LLM-RAG retrieval, and speaker embeddings in a rule-based risk score to flag narrative reuse, structural inconsistencies, and cross-case voice repetition while balancing sensitivity and false positives.

What carries the argument

The rule-based risk score that fuses NER, regex features, LLM-RAG retrieval, and speaker embeddings to detect narrative reuse, inconsistencies, and cross-case voice repetition in synthetic multimodal data.

If this is right

  • The framework supports detection of fraud indicators through combined linguistic, structural, and acoustic signals without needing private real-world datasets.
  • Rule-based scoring balances sensitivity to fraud against false positive rates in simulated conditions.
  • Dataset validation and component evaluations demonstrate stability and potential transfer to operational settings.
  • It supplies a reproducible baseline for multimodal fraud detection that extends past text-only approaches.

Where Pith is reading between the lines

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

  • If the synthetic data matches real interactions closely enough, the pipeline could support automated initial screening that reduces manual claim reviews.
  • The generation and scoring approach might adapt to other conversational domains such as banking disputes or customer service abuse detection.
  • Speaker embeddings open the possibility of linking related claims across different organizations if privacy safeguards allow voice data sharing.

Load-bearing premise

The synthetic data generation process accurately replicates the linguistic, behavioural, and acoustic properties of real FNOL agent-customer interactions, allowing the rule-based risk score to generalize to actual fraud detection.

What would settle it

Testing the full pipeline on a collection of real FNOL audio recordings with known fraud labels and measuring how well the risk scores align with those labels in precision and recall.

Figures

Figures reproduced from arXiv: 2606.28002 by Abdul Hamid Sadka, Amal Htait, Emma Meisingseth, Karishma Jaitly, Muhammad Shakeel Akram.

Figure 1
Figure 1. Figure 1: Proposed end-to-end solution for AI-based robust insurance fraud claims detection. 3.1. Synthetic Dataset Generation [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: illustrates the synthetic data generation process. The pipeline begins with user-specified inputs, including (i) fraudu￾lent and legitimate dialogue templates, and (ii) structured vari￾ables such as customer name, age, postcode, policy number, and product type. These inputs are passed to a GPT-2 transformer model, which generates diverse and coherent synthetic tran￾scripts. Real FNOL calls often include sm… view at source ↗
Figure 3
Figure 3. Figure 3: Block diagram illustrating the fraud risk scoring framework. 3.6.1. Structured feature extraction. Customer histories are retrieved using extracted entities (policy number, name, postcode). From these records, the following risk indicators are computed: 1. Early claims: whether a claim is filed within 30 days of pol￾icy inception. 2. High-cost claims: whether the claim amount exceeds 1.2× the average claim… view at source ↗
Figure 4
Figure 4. Figure 4: Feature extraction from xTTS based customer-agent recordings. note that the WER is computed using STT transcriptions of fully synthetic audio generated from synthetic text. The results un￾der these conditions highlight the effectiveness of the proposed workflow and its potential for further improvement with even more enhanced synthesis quality. 5.3. Binary Classification The BERT–RAG classifier was trained… view at source ↗
Figure 5
Figure 5. Figure 5: Cluster formation on Common Voice samples (39 unique speakers, 136 samples) [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Binary classification applied to claim-portal–style customer transcripts (GPT-2) and agent–customer call (xTTS) scenarios [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Voice clustering of xTTS-based agent–customer recordings. 5.4. Voice Clustering Speaker embeddings derived from diarisation timestamps are used to isolate customer segments for verification and cluster￾ing. This enables detection of repeated speakers across claims, an indicator of potential organised fraud [PITH_FULL_IMAGE:figures/full_fig_p007_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Clusters overall detected for common voice recordings based on voice embedding; in addition to [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗
read the original abstract

Insurance fraud imposes substantial financial losses and operational inefficiencies, raising premiums and impacting trust among legitimate policyholders. Early detection at FNOL remains a persistent challenge. Existing approaches rely largely on private, text-only datasets, limiting progress on multimodal methods that integrate linguistic, behavioural, and speaker-based indicators. We introduce a synthetic multimodal framework that replicates FNOL conditions. It generates agent-customer dialogue transcripts and two-speaker audios, performs ASR and diarisation. Downstream modules combine NER, regex-based feature extraction, LLM-RAG retrieval, and speaker embeddings in a rule-based risk score to flag narrative reuse, structural inconsistencies, and cross-case voice repetition while balancing sensitivity and false positives. Dataset validation and component-level evaluations show stability and transfer potential, offering a reproducible baseline beyond text-only fraud detection.

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 / 1 minor

Summary. The manuscript introduces a synthetic multimodal framework for insurance fraud detection at the First Notice of Loss (FNOL) stage. It generates agent-customer dialogue transcripts and two-speaker audios, performs ASR and diarization, then applies NER, regex-based feature extraction, LLM-RAG retrieval, and speaker embeddings within a rule-based risk score to identify indicators such as narrative reuse, structural inconsistencies, and cross-case voice repetition. The authors state that dataset validation and component-level evaluations demonstrate stability and transfer potential, positioning the work as a reproducible baseline beyond text-only approaches.

Significance. If the synthetic data generation accurately captures real FNOL linguistic, behavioral, and acoustic properties and the rule-based score generalizes, the pipeline could provide a valuable public benchmark for multimodal fraud detection, addressing the reliance on private datasets. The integration of multiple modalities and hybrid components represents a constructive direction, but the significance is currently limited by the absence of quantitative validation.

major comments (2)
  1. [Abstract] Abstract: The claim that 'dataset validation and component-level evaluations show stability and transfer potential' is unsupported by any reported metrics (e.g., precision, recall, F1 for the risk score; diarization error rates; distributional similarity between synthetic and real transcripts on fraud indicators like narrative reuse). Without these, the central claim of a functional pipeline cannot be evaluated.
  2. [Methods / Risk Score Module] Pipeline and risk score description: The rule-based risk score depends on unspecified free parameters (thresholds and feature weights) whose selection process is not described, nor is any sensitivity analysis or justification provided for balancing sensitivity against false positives. This directly affects reproducibility and the validity of the fraud flagging claims.
minor comments (1)
  1. [Abstract] Abstract: The phrasing 'two-speaker audios' is imprecise; consider 'two-speaker audio recordings' or 'binaural audio' for clarity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback highlighting the need for quantitative metrics and greater reproducibility in the risk score. We address each major comment below and will revise the manuscript to strengthen these aspects.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The claim that 'dataset validation and component-level evaluations show stability and transfer potential' is unsupported by any reported metrics (e.g., precision, recall, F1 for the risk score; diarization error rates; distributional similarity between synthetic and real transcripts on fraud indicators like narrative reuse). Without these, the central claim of a functional pipeline cannot be evaluated.

    Authors: We agree the abstract claim is not supported by explicit metrics in the current version. The evaluations performed were primarily qualitative and stability checks on synthetic data generation rather than full quantitative benchmarking against real FNOL data. In revision we will remove or qualify the claim in the abstract and add a dedicated results section reporting component-level metrics (e.g., ASR WER, diarization DER, NER F1, and risk-score precision/recall on held-out synthetic cases) along with distributional comparisons where feasible. revision: yes

  2. Referee: [Methods / Risk Score Module] Pipeline and risk score description: The rule-based risk score depends on unspecified free parameters (thresholds and feature weights) whose selection process is not described, nor is any sensitivity analysis or justification provided for balancing sensitivity against false positives. This directly affects reproducibility and the validity of the fraud flagging claims.

    Authors: We acknowledge the parameters and their selection process are under-specified. The current rule-based score uses fixed thresholds derived from pilot runs on the synthetic corpus and simple equal weighting of the three indicator categories, but these details and any sensitivity testing were omitted. In the revised manuscript we will document the exact thresholds and weights, explain their derivation from observed feature distributions, and add a sensitivity analysis varying each parameter by ±20% to show impact on false-positive rate. revision: yes

Circularity Check

0 steps flagged

No circularity: synthetic pipeline and rule-based score are self-contained descriptions

full rationale

The paper introduces a synthetic data generation process, ASR/diarisation, NER/regex/LLM-RAG/speaker-embedding modules, and a rule-based risk score as a new framework. No equations, fitted parameters renamed as predictions, self-definitional loops, or load-bearing self-citations appear in the provided abstract or described components. The central claim rests on the synthetic replication assumption and component evaluations, which are presented as external validation rather than reductions to prior inputs by construction. This is the common honest non-finding for descriptive pipeline papers.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the untested assumption that synthetic data can stand in for real FNOL interactions and that a rule-based combination of the listed features will reliably surface fraud signals.

free parameters (1)
  • risk score thresholds and feature weights
    The rule-based risk score must combine multiple heterogeneous signals and therefore requires tuned cutoffs or weights to balance sensitivity against false positives; these are not specified in the abstract.
axioms (1)
  • domain assumption Synthetic generation of agent-customer dialogues and two-speaker audio can replicate the statistical properties of real FNOL calls sufficiently for downstream fraud detection
    The entire pipeline is constructed on top of this synthetic data; if the distribution diverges from reality the flagged indicators lose meaning.

pith-pipeline@v0.9.1-grok · 5682 in / 1412 out tokens · 63086 ms · 2026-06-29T04:25:57.327972+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

62 extracted references · 3 canonical work pages · 1 internal anchor

  1. [1]

    Dialogue to Detection: A Multimodal Hybrid NLP Pipeline for Insurance Fraud Detection

    Introduction Insurance fraud is a persistent and high-cost problem world- wide, with estimates putting annual losses over $300 billion in the US and more than £1.1 billion in the UK [1, 2, 3, 4, 5]. Such losses affect not only insurers, but also legitimate policy- holders through increased premiums and reduced benefits. The claim process, especially at th...

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  2. [2]

    model statements

    Background and Related Work Multimodal data, AI, and NLP enablescross-modal verifica- tion, linking voice identity with narrative similarity for fraud detection, but their public availability remains scarce, particu- larly in the insurance domain [10, 6]. This scarcity is a recurring limitation in the literature and motivates synthetic multimodal datasets...

  3. [3]

    remain text-only and lack conversational speech, leaving a gap for multimodal synthetic dataset

  4. [4]

    um,” “uh,

    Proposed End-to-End Pipeline As outlined, existing works on insurance fraud detection is hin- dered by the absence of publicly shareable multimodal datasets, constrained by privacy and regulatory requirements. This lim- its progress even as fraudulent claims impose substantial finan- cial losses and drive higher premiums. To address this gap, we present (...

    2024

  5. [5]

    Experimental Setup The pipeline is implemented entirely with open-source frame- works to ensure reproducibility. Core components from Hug- ging Face Transformers [40] provide pre-trained BERT [7] and GPT-2 models [41, 8], reducing training cost and en- abling domain adaptation for insurance narratives. GPT-2 (GPT2LMHeadModel), paired with its tokeniser, g...

  6. [6]

    Sentence-Transformer embeddings (all- MiniLM-L6-v2 [47]) are indexed in FAISS [48, 49] for fast se- mantic retrieval, enabling RAG [50] via LangChain [51]

    with regex rules to capture structured identifiers (e.g., policy numbers, postcodes) and support linkage to simulated customer histories. Sentence-Transformer embeddings (all- MiniLM-L6-v2 [47]) are indexed in FAISS [48, 49] for fast se- mantic retrieval, enabling RAG [50] via LangChain [51]. BERT is fine-tuned for binary fraud classification, leveraging ...

  7. [7]

    Results and Discussion We present the results associated with each of the components in the end-to-end pipeline, demonstrating the system architec- ture’s plausibility. 5.1. Generated Synthetic Datasets The generated dataset is designed to balance fraudulent and le- gitimate claims while reflecting the multimodal nature of FNOL data. In total, it comprise...

  8. [8]

    This work presents a proof-of-function pipeline that unifies synthetic data genera- tion, speech processing, and fraud detection within a risk scor- ing architecture

    Conclusion We present a synthetic multimodal framework for insurance fraud detection, addressing the absence of shareable audio–text datasets constrained by privacy regulations. This work presents a proof-of-function pipeline that unifies synthetic data genera- tion, speech processing, and fraud detection within a risk scor- ing architecture. The system i...

  9. [9]

    Major new crackdown on insurance fraud,

    Home Office and The Rt Hon Lord Hanson of Flint, “Major new crackdown on insurance fraud,” https://www.gov.uk/government/ news/major-new-crackdown-on-insurance-fraud, October 2024, accessed August 2025

    2024

  10. [10]

    Ifb’s 2024 annual report has been published,

    Insurance Fraud Bureau, “Ifb’s 2024 annual report has been published,” https://www.insurancefraudbureau.org/media-centre/ ifb-news/2025/ifb-s-2024-annual-report-has-been-published, 2025, accessed August 2025

    2024

  11. [11]

    Insur- ance topics — insurance fraud,

    National Association of Insurance Commissioners, “Insur- ance topics — insurance fraud,” https://content.naic.org/ insurance-topics/insurance-fraud, 2024, accessed August 2025

    2024

  12. [12]

    In- surance fraud detection: Evidence from artificial intelligence and machine learning,

    F. Aslam, A. I. Hunjra, Z. Ftiti, W. Louhichi, and T. Shams, “In- surance fraud detection: Evidence from artificial intelligence and machine learning,”Technological Forecasting and Social Change, 2022

    2022

  13. [13]

    Financial fraud detection based on machine learning: A systematic literature re- view,

    A. Ali, S. A. Razak, S. H. Othman, T. A. E. Eisa, A. Al-Dhaqm, M. Nasser, T. Elhassan, H. Elshafie, and A. Saif, “Financial fraud detection based on machine learning: A systematic literature re- view,”IEEE Access, 2022

    2022

  14. [14]

    Detection of insurance fraud us- ing nlp and ml: A study on three different nlp-techniques for text classification,

    R. B ¨acklund and H. ¨Ohman, “Detection of insurance fraud us- ing nlp and ml: A study on three different nlp-techniques for text classification,” Master’s thesis, Lund University, 2023

    2023

  15. [15]

    Bert: Pre- training of deep bidirectional transformers for language under- standing,

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “Bert: Pre- training of deep bidirectional transformers for language under- standing,” inProceedings of the 2019 conference of the North American chapter of the association for computational linguis- tics: human language technologies, volume 1 (long and short pa- pers), 2019, pp. 4171–4186

    2019

  16. [16]

    Gpt understands, too,

    X. Liu, Y . Zheng, Z. Du, M. Ding, Y . Qian, Z. Yang, and J. Tang, “Gpt understands, too,”AI Open, vol. 5, pp. 208–215, 2024

    2024

  17. [17]

    Xlnet: Generalized autoregressive pretraining for lan- guage understanding,

    Z. Yang, Z. Dai, Y . Yang, J. Carbonell, R. R. Salakhutdinov, and Q. V . Le, “Xlnet: Generalized autoregressive pretraining for lan- guage understanding,”Advances in neural information processing systems, vol. 32, 2019

    2019

  18. [18]

    Classification of transcribed voice recordings: Deter- mining the claim type of recordings submitted by swedish insur- ance clients,

    C. Piehl, “Classification of transcribed voice recordings: Deter- mining the claim type of recordings submitted by swedish insur- ance clients,” Master’s thesis, KTH Royal Institute of Technology, 2021

    2021

  19. [19]

    Insurance claim fraud dataset,

    A. Abdelrahim and M. Data, “Insurance claim fraud dataset,” 2020, https://data.mendeley.com/datasets/992mh7dk9y/1

    2020

  20. [20]

    92k real-world call center scripts (en- glish),

    AIxBlock, “92k real-world call center scripts (en- glish),” https://huggingface.co/datasets/AIxBlock/ 92k-real-world-call-center-scripts-english, 2024, accessed August 2025

    2024

  21. [21]

    You cannot hide your telephone lies: Providing a model statement as an aid to detect deception in insurance telephone calls,

    S. Leal, A. Vrij, L. Warmelink, Z. Vernham, and R. P. Fisher, “You cannot hide your telephone lies: Providing a model statement as an aid to detect deception in insurance telephone calls,”Applied Cognitive Psychology, vol. 32, no. 6, 2018

    2018

  22. [22]

    Fraud de- tection in telephone conversations for financial services using lin- guistic features,

    N. Bajaj, T. G. Constance, M. Rajwadi, J. Wall, M. Moniri, C. Glackin, N. Cannings, C. Woodruff, and J. Laird, “Fraud de- tection in telephone conversations for financial services using lin- guistic features,” inProceedings of Interspeech, 2020

    2020

  23. [23]

    Practical guideline to efficiently detect insurance fraud in the era of machine learn- ing: A household insurance case,

    D. Banulescu-Radu and M. Yankol-Schalck, “Practical guideline to efficiently detect insurance fraud in the era of machine learn- ing: A household insurance case,”Journal of Risk and Insurance, vol. 91, no. 4, pp. 867–913, 2024

    2024

  24. [24]

    Design of a nlp-empowered finance fraud awareness model: the anti-fraud chatbot for fraud detection and fraud classification as an instance,

    J.-W. Chang, N. Yen, and J. C. Hung, “Design of a nlp-empowered finance fraud awareness model: the anti-fraud chatbot for fraud detection and fraud classification as an instance,”Journal of Am- bient Intelligence and Humanized Computing, vol. 13, no. 10, pp. 4663–4679, 2022

    2022

  25. [25]

    Design of a nlp-empowered finance fraud awareness model: the anti-fraud chatbot for fraud detection and fraud classi- fication as an instance,

    ——, “Design of a nlp-empowered finance fraud awareness model: the anti-fraud chatbot for fraud detection and fraud classi- fication as an instance,”Journal of Ambient Intelligence and Hu- manized Computing, 2023

    2023

  26. [26]

    En- hancing claims handling processes with insurance based language models,

    A. Dimri, S. Yerramilli, P. Lee, S. Afra, and A. Jakubowski, “En- hancing claims handling processes with insurance based language models,” inProceedings of the AAAI Workshop on AI in Insur- ance, 2024

    2024

  27. [27]

    Ai in insurance: Enhancing fraud detection and risk assessment,

    C. M. Gangani, “Ai in insurance: Enhancing fraud detection and risk assessment,”International Journal of Advanced Computer Science and Applications, 2023

    2023

  28. [28]

    Ai in fraud detection: Leveraging machine learning to combat insurance fraud,

    V . K. Tarra, “Ai in fraud detection: Leveraging machine learning to combat insurance fraud,”International Journal of Innovative Technology and Exploring Engineering, 2024

    2024

  29. [29]

    Innovative applications of ai and machine learn- ing in fraud detection for insurance claims,

    R. A. Perumal, “Innovative applications of ai and machine learn- ing in fraud detection for insurance claims,”Journal of Risk and Financial Management, 2023

    2023

  30. [30]

    Practical guideline to efficiently detect insurance fraud in the era of machine learning: A household insurance case,

    D. Banulescu-Radu and M. Yankol-Schalck, “Practical guideline to efficiently detect insurance fraud in the era of machine learning: A household insurance case,”Journal of Risk Finance, 2023

    2023

  31. [31]

    Detecting fraud calls vis- `a-vis natural language processing,

    P. K. Kumar, S. Ray, L. Kumarasankaralingam, A. Ramamoorthy, P. Kumar, and A. Dutta, “Detecting fraud calls vis- `a-vis natural language processing,” inProceedings of the International Confer- ence on Computer Communication and Informatics, 2022

    2022

  32. [32]

    Detection of spam and fraudulent calls using natural language processing model,

    A. Gupta, “Detection of spam and fraudulent calls using natural language processing model,”International Journal of Computer Applications, 2024

    2024

  33. [33]

    Multimodal detection framework for financial fraud integrating llms and interpretable machine learning,

    H. Nie, Z. Long, Z. Fang, and L. Gao, “Multimodal detection framework for financial fraud integrating llms and interpretable machine learning,”Journal of Data and Information Science, vol. 10, no. 4, pp. 1–25, 2025

    2025

  34. [34]

    Fraud de- tection in healthcare claims using machine learning: A systematic review,

    A. du Preez, S. Bhattachary, P. Beling, and E. Bowen, “Fraud de- tection in healthcare claims using machine learning: A systematic review,”Health Informatics Journal, 2023

    2023

  35. [35]

    Auto in- surance fraud detection with multimodal learning,

    J. Yang, K. Chen, K. Ding, C. Na, and M. Wang, “Auto in- surance fraud detection with multimodal learning,”Data Intelli- gence, vol. 5, no. 2, pp. 388–412, 2023

    2023

  36. [36]

    Autofraudnet: A multimodal network to detect fraud in the auto insurance industry,

    A. Asgarian, R. Saha, D. Jakubovitz, and J. Peyre, “Autofraud- net: A multimodal network to detect fraud in the auto insurance industry,”arXiv preprint arXiv:2301.07526, 2023

    work page arXiv 2023

  37. [37]

    How generative ai is transforming insurance claims: Inside swiss re’s claimsgenai,

    F. Maurer and V . Plantard, “How generative ai is transforming insurance claims: Inside swiss re’s claimsgenai,” 2025, https://www.swissre.com/ risk-knowledge/advancing-societal-benefits-digitalisation/ how-generative-ai-is-transforming-insurance-claims-claimsgenai. html

    2025

  38. [38]

    Enterprise fraud detection systems — custom ai solu- tions,

    Xenoss, “Enterprise fraud detection systems — custom ai solu- tions,” 2025, https://xenoss.io/solutions/fraud-detection

    2025

  39. [39]

    Fraud risk management — moody’s data & ana- lytics solutions,

    M. Analytics, “Fraud risk management — moody’s data & ana- lytics solutions,” 2025, https://www.moodys.com/web/en/us/kyc/ solutions/fraud-prevention.html

    2025

  40. [40]

    E-commerce fraud detection based on machine learning techniques: Systematic literature review,

    A. Mutemi and F. Bacao, “E-commerce fraud detection based on machine learning techniques: Systematic literature review,”Jour- nal of Information Security and Applications, 2023

    2023

  41. [41]

    Fraud detection with natural language processing,

    P. Boulieris, J. Pavlopoulos, A. Xenos, and V . Vassalos, “Fraud detection with natural language processing,”Machine Learning, 2024, https://github.com/pboulieris/FraudNLP

    2024

  42. [42]

    Insurance-claim-fraud-detection (github repository),

    Nirab, “Insurance-claim-fraud-detection (github repository),” 2021, https://github.com/nirab25/ Insurance-Claim-Fraud-Detection

    2021

  43. [43]

    Insurance claims fraud detection model (github repository),

    P. S. Prakash, “Insurance claims fraud detection model (github repository),” 2023, https://github.com/PrajwalSuryaPrakash/ Insurance-Claims-Fraud-Detection-Model

    2023

  44. [44]

    Predictive analysis and fraud detection for insurance claims (github pages),

    M. Gaikwad, “Predictive analysis and fraud detection for insurance claims (github pages),” 2023, https://manojgaikwad13.github.io/ Predictive-Analysis-and-Fraud-Detection-for-Insurance-Claims/

    2023

  45. [45]

    Fraud detector (github repository),

    Y . Cheuk, “Fraud detector (github repository),” 2023, https:// github.com/yingcheuk/fraud-detector

    2023

  46. [46]

    A complete guide to natural lan- guage processing,

    DeepLearning.AI, “A complete guide to natural lan- guage processing,” https://www.deeplearning.ai/resources/ natural-language-processing/, 2025, accessed: 17-July-2025

    2025

  47. [47]

    Families and households,

    Office for National Statistics, “Families and households,” 2025, accessed: 2025-08-29. [Online]. Available: https://www.ons.gov. uk/peoplepopulationandcommunity/birthsdeathsandmarriages/ families/datasets/familiesandhouseholdsfamiliesandhouseholds

    2025

  48. [48]

    Transformers: State-of-the-art natural language processing,

    T. Wolf, L. Debut, V . Sanh, J. Chaumond, C. Delangue, A. Moi, P. Cistac, T. Rault, R. Louf, M. Funtowicz, J. Davison, S. Shleifer, P. von Platen, C. Ma, Y . Jernite, J. Plu, C. Xu, T. L. Scao, S. Gugger, M. Drame, Q. Lhoest, and A. M. Rush, “Transformers: State-of-the-art natural language processing,” inProceedings of the 2020 Conference on Empirical Met...

    2020

  49. [49]

    Language models are unsupervised multitask learn- ers,

    A. Radford, J. Wu, R. Child, D. Luan, D. Amodei, and I. Sutskever, “Language models are unsupervised multitask learn- ers,” 2019, openAI

    2019

  50. [50]

    Coqui tts/xtts: A deep learning toolkit for high-quality text-to-speech,

    C. AI, “Coqui tts/xtts: A deep learning toolkit for high-quality text-to-speech,” https://github.com/coqui-ai/TTS, 2023

    2023

  51. [51]

    XTTS: A massively multilingual zero-shot text-to-speech model,

    E. Casanova, K. Davis, E. G ¨olge, G. G ¨oknar, I. Gulea, L. Hart, A. Aljafari, J. Meyer, R. Morais, S. Olayemiet al., “Xtts: a massively multilingual zero-shot text-to-speech model,”arXiv preprint arXiv:2406.04904, 2024

    work page arXiv 2024

  52. [52]

    gtts: Google text-to-speech python library,

    D. Pellegrini, “gtts: Google text-to-speech python library,” https: //gtts.readthedocs.io/, 2024

    2024

  53. [53]

    Whisperx: Time-accurate speech transcription of long-form audio,

    M. Bain, J. Huh, T. Han, and A. Zisserman, “Whisperx: Time-accurate speech transcription of long-form audio,”INTER- SPEECH 2023, 2023

    2023

  54. [54]

    Jean-baptiste/roberta-large-ner-english,

    J.-B. Polle, “Jean-baptiste/roberta-large-ner-english,” https://huggingface.co/Jean-Baptiste/roberta-large-ner-english, 2021, fine-tuned RoBERTa model for Named Entity Recognition on CoNLL-2003 dataset

    2021

  55. [55]

    all-minilm-l6-v2,

    Sentence-Transformers, “all-minilm-l6-v2,” accessed: 2025- 09-05. [Online]. Available: https://huggingface.co/ sentence-transformers/all-MiniLM-L6-v2

    2025

  56. [56]

    The faiss library,

    M. Douze, A. Guzhva, C. Deng, J. Johnson, G. Szilvasy, P.-E. Mazar´e, M. Lomeli, L. Hosseini, and H. J´egou, “The faiss library,” arXiv, 2024

    2024

  57. [57]

    Billion-scale similarity search with GPUs,

    J. Johnson, M. Douze, and H. J ´egou, “Billion-scale similarity search with GPUs,”IEEE Transactions on Big Data, vol. 7, no. 3, pp. 535–547, 2019

    2019

  58. [58]

    Retrieval- augmented generation for knowledge-intensive nlp tasks,

    P. Lewis, E. Perez, A. Piktus, F. Petroni, V . Karpukhin, N. Goyal, H. K¨uttler, M. Lewis, W.-t. Yih, T. Rockt¨aschelet al., “Retrieval- augmented generation for knowledge-intensive nlp tasks,”Ad- vances in neural information processing systems, vol. 33, pp. 9459–9474, 2020

    2020

  59. [59]

    Langchain,

    LangChain, “Langchain,” https://github.com/langchain-ai/ langchain, 2025, accessed: 17-July-2025

    2025

  60. [60]

    Resemblyzer: V oice em- beddings for speaker similarity,

    H. W. Corentin Jemineet al., “Resemblyzer: V oice em- beddings for speaker similarity,” https://github.com/resemble-ai/ Resemblyzer, 2020

    2020

  61. [61]

    Scikit-learn: Machine learning in Python,

    F. Pedregosa, G. Varoquaux, A. Gramfort, V . Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V . Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, and E. Duchesnay, “Scikit-learn: Machine learning in Python,” Journal of Machine Learning Research, vol. 12, pp. 2825–2830, 2011

    2011

  62. [62]

    Common voice,

    Mozilla, “Common voice,” https://commonvoice.mozilla.org/ pcm/datasets, 2024, accessed: 2025-09-01

    2024