Skip to main content

Real-time Chatbot Romance Scam Detection in Stage Classification

Yi-cheng Liu ; Guan-ru Lai ; Yi-cheng Lai ; Lyn Chao-ling Chen (2026) — arXiv (Cornell University)

Source monitoring coverage

Checked Oct 30, 2026 19:15 UTC — no notice found.

DataCite identity and deposited events checked. DataCite incoming relationships checked. arXiv submission history checked. No explicit source-update notice identified in the checked information. Coverage is incomplete and does not certify source validity. Routine version changes do not trigger writing.

Citation activity

Total citations · Google Scholar
Unavailable
Average per calendar year
Citation count unavailable

Counts reflect Google Scholar’s coverage and do not measure research quality. Source, calculation and limitations

View Original Source (opens in a new tab)

DOI: 10.48550/arxiv.2610.20631 (opens in a new tab)

Country:   Taiwan
Tags:   Gpt 4O, Line Messaging Api, Prevention Advice, Real Time Chatbot, Reddit Dataset, Romance Scam Detection, Stage Classification, Whitty Stage Model
What do these research terms mean?
Preprint
A manuscript shared before formal peer review and publication. Check whether a later published version is available.
Dataset
A collection of data or examples for others to inspect or reuse. It can appear in Library search and topic mapping, but RSRC does not use dataset records as evidence in Research Insights.
Dissertation or thesis
Research submitted for an academic degree. This describes its format, not its reliability.
Journal article
An article published in a journal. This label alone does not establish peer review, study quality, or how well the findings apply elsewhere.
Qualitative research
Examines experiences, meanings, or processes, often through interviews or observations. It can explain how something happens without estimating how common it is.
Quantitative research
Uses numerical measurements to describe patterns or test relationships. A relationship between two measurements does not by itself show that one causes the other.
Systematic review
Uses a planned, documented method to find and assess research addressing a question. Its conclusions still depend on the included studies and what the search covered.
Meta-analysis
Statistically combines results from multiple studies. Combining studies does not remove weaknesses in their design or make unlike populations interchangeable.
Not classified
This record has no recognized label in this filter. It does not mean the publication used no method, or that no research exists.

Definitions draw on DataCite resource types; Cochrane review methods; NLM: association and causation. RSRC’s dataset and classification rules are explained in our methodology.

Citation tools
Download RIS Download BibTeX 


              
              

            

          

        


      

      

    

        
Transparency

        
Evidence and review status

        
This page contains AI-generated content. No human content review or subject-matter-expert review is recorded.

    


    

                    

                
Source basis

                

                    Downloaded PDF
                

            

                    

                
Source updates

                

                    No notice found at last check
                

            

                    

                
AI-generated page content

                

                    Yes
                

            

                    

                
Automated checks

                

                    Passed
                

            

                    

                
Administrative approval

                

                    Yes
                

            

                    

                
Human content review

                

                    Not recorded
                

            

                    

                
Subject-matter-expert review

                

                    Not recorded
                

            

            


    

        
            How this was prepared
        
        

                            

                    
Source basis

                    

                    
RSRC downloaded and privately stored a copy of the paper for internal analysis. The PDF is not offered to viewers from this page.


                                            
    
                                                            
  • 
                                    PDF added to RSRC:
                                    
                                
    • 
                                                    

                    
                                        

                

                            

                    
Source updates

                    

                    
No incoming update notice was found in the dated Crossref response. Coverage is incomplete, particularly for corrections and expressions of concern; this is not a guarantee that the source is valid or unchanged.


                                            
    
                                                            
  • 
                                    Last source-status attempt:
                                    
                                
    • 
                                                    

                    
                                        

                

                            

                    
AI-generated page content

                    

                    
AI-generated research notes displayed on this page: Synopsis, Identified gaps, Methods, Limitations, Future work. The paper itself is not described as AI-generated.


                                            
    
                                                            
  • 
                                    Document analysis recorded:
                                    
                                
    • 
                                                            
  • 
                                    Synopsis generation recorded:
                                    
                                
    • 
                                                            
  • 
                                    Page record updated:
                                    
                                
    • 
                                                    

                    
                                        

                

                            

                    
Automated checks

                    

                    
The current, source-bound synopsis passed the recorded versioned publication checks.


                                            
    
                                                            
  • 
                                    Checks completed:
                                    
                                
    • 
                                                    

                    
                                            

                            
                                View passed checks (3)
                            
                            
    
                                                                    
  • Length, completeness, repetition, refusal, boilerplate, and active-markup screening
    • 
                                                                    
  • Numerical claims checked against the available source text
    • 
                                                                    
  • English-source lexical grounding check
    • 
                                                            

                        

                                        

                

                            

                    
Administrative approval

                    

                    
An authenticated administrator approved the bibliographic record for public Library display. This is not a review of every research claim.


                                            
    
                                                            
  • 
                                    Approved for public display:
                                    
                                
    • 
                                                    

                    
                                        

                

                            

                    
Human content review

                    

                    
No human review is recorded for the AI-generated content displayed on this page.


                    
                                        

                

                            

                    
Subject-matter-expert review

                    

                    
RSRC has not recorded review of this content by a subject-matter or methods expert.


                    
                                        

                

                    

    


    

        Review-state definitions

        Found a possible error?
        Request a correction.
    



    



              

        
                  

            
Synopsis

            
The publication reports on a real‑time romance scam detection system embedded in a Line messaging chatbot. The authors frame the problem as a social‑scientific seven‑stage process and implement a practical application that analyzes user messages to identify the stage of a romance scam and provide explanations and prevention tips. The system architecture combines an NLP pipeline with a large language model to perform context analysis, stage classification, and user guidance, delivering stage names, brief descriptions, and stage‑specific prevention advice via the chat interface. Methodologically, the researchers adopt Whitty’s seven‑stage framework and use GPT‑4o to classify Line messages into stages ranging from motivated partner search to re‑victimization. The model also extracts relevant scam-related labels and supplies concise rationales for classifications. Explanations and prevention content are generated in a dedicated module, with the prevention module offering stage‑appropriate guidance. A Line webhook component enables real‑time processing of user input and output of explanations or prevention tips, enabling ongoing user interaction during a chat session. In evaluating the approach, the authors describe a dataset of romance scam data collected from Reddit (n=152) labeled with seven stages and report stage‑level performance metrics. They note better performance for some stages (e.g., continuation of the scam and re‑victimization) and challenges for others (notably the sting stage when data consist of single messages). The work suggests the system can function in real‑world Line chats to flag risk and educate users about the scam process, while acknowledging that performance varies by stage and is limited by available contextual information.

          

        
                  

            
Identified Gaps

            
The evaluation does not directly validate the system's central claim that conversation context improves stage detection: reported results are explicitly for individual stage samples without context, while better real-world performance is asserted rather than measured. The paper also provides no baseline comparison, inter-rater labeling reliability, participant study, or assessment of prevention outcomes. Stage-level evidence is uneven, with only four sexual-abuse samples and poor sting-stage accuracy.

          

        
                  

            
Methods

            
The authors built a Python Flask and Line Messaging API chatbot that sends user message histories to GPT-4o. Guided by Whitty's seven-stage romance-scam framework, a prompt classifies input type, scam stage, and trigger labels, then generates explanations and three prevention suggestions. Evaluation used 152 Reddit-sourced romance-scam texts manually labeled into seven stages and reported per-stage accuracy for classification without context information.

          

        
                  

            
Limitations

            
The reported evaluation uses 152 manually labeled Reddit texts and tests individual stage samples without context, although the deployed system relies on message-history context. Performance varies greatly by stage, from 28.6% for the sting stage to 100% for sexual abuse, but the latter has only four samples. The paper does not report annotation procedures or agreement, overall accuracy, baseline comparisons, error metrics, or a real-world deployment evaluation.

          

        
                  

            
Future Work

            
Future research should test the chatbot on complete, real-time conversation histories and compare context-aware performance with the reported single-stage, no-context evaluation. It should also use larger, more balanced, independently labeled datasets, particularly for the low-performing sting stage and sparsely represented sexual-abuse stage. User-centered studies should assess whether explanations, visual feedback, and stage-specific advice reduce victimization without causing harmful false alarms.

          

              

    
    

        
Related Research Insights

        

            See how this publication connects to RSRC's living evidence syntheses through current citations and research-topic mapping.