Conversational Artifical Intelligence - a unique real-time voice analytics technology providing highly accurate behavioural determinations - scalably enabling the identification of fraud in insurance claims

会话人工智能 - 一种独特的实时语音分析技术，可提供高度准确的行为确定 - 可扩展地识别保险索赔中的欺诈行为

• 批准号: 104997
• 金额: $ 35.64万
• 依托单位: CONVERSATIONAL AI GROUP LIMITED
• 依托单位国家: 英国
• 项目类别: Feasibility Studies
• 财政年份: 2019
• 资助国家: 英国
• 起止时间: 2019 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=104997
• 关键词: Conversational; Artifical; Intelligence; unique; real

"Latest figures for 2017 indicate the volume of frauds recorded by Action Fraud is the highest it has ever been at 272,980 offences(5,250/week), insurance fraud totals a value of £3B/yr. Dishonest liability claims (""slip and trip"") rose 36%/yr since 2015 with claims value up 14% (£391M), however the biggest victim is car insurance, with 80,000 claims detected 2017(£1B). These resulted in increases of insurance premiums and time/review before insurers pay legitimate claims (increased scrutiny of honest customers).At insurance contact centres (first contact point), there are limited current solutions to identify exaggerated claims information. Statistical models identify claim types that have higher probability of being fraudulent; voice stress tests identify changes in pitch&tone. These don't provide a defined output (only indicate 'likelihood') and are easily influenced by line of conversation/first languages/mental health/hearing difficulties. No current solution can measure linguistical changes, which is the primary identification parameter for this type of fraud.Conversational AI (CAI) is developing an entirely innovative model combining the unique team expertise of insurance fraud/behavioural analytics/detective work/machine learning/language processing to a near real-time call ""credibility"" score to identify fraudulent insurance claims. The company plans to provide their software as an add-on service to existing call recording platforms thereby offering: automated screening of every call increasing investigation efficiency by a factor of 60/data transparency/enhanced customer outcomes/safeguarding vulnerable customers/ensuring regulatory compliance/improving anti-fraud measures.CAI are well placed to exploit this opportunity, having already established basic linear algorithms with the ability to detect 50 nuances in human behaviour to understand contextualise emotion, successfully identifying deception using manually transcribed call data. As a new joint venture combining Intelligent Voice's existing and proven language processing capability with Strenuus' behavioural analysis expertise, CAI will develop a software system capable of identifying human interaction using GPU powered automatic speech recognition transcripts. Interest has already been expressed from voice biometric software providers offering open access to their API for data acquisition, and contact centres who have confirmed their willingness to trial the solution.With support from Innovate UK a 12-month programme of R&D is required to deliver a validated system in a closed live environment, identifying crucial markets in caller credibility assessment based on-agent and caller interaction. CAI has the potential to disrupt the voice biometric and anti-fraud markets with global exploitation potential. The project will deliver export led growth for CAI, a substantial ROI (12700%), increased employment(25 additional jobs) and further R&D opportunities."

“2017年的最新数据显示，Action Fraud记录的欺诈数量是有史以来最高的，达到272，980起犯罪（每周5，250起），保险欺诈的总价值为每年30亿英镑。自2015年以来，不诚实责任索赔（“滑倒和绊倒”）每年增长36%，索赔价值增长14%（3.91亿英镑），但最大的受害者是汽车保险，2017年发现了8万起索赔（10亿英镑）。这导致保险公司支付合法索赔之前的保险费和时间/审查的增加（对诚实客户的审查增加）。在保险联络中心（第一联络点），目前识别夸大索赔信息的解决方案有限。统计模型识别具有较高欺诈概率的索赔类型;语音压力测试识别音高和音调的变化。这些不提供定义的输出（仅指示“可能性”），并且很容易受到对话线/第一语言/心理健康/听力困难的影响。会话人工智能（CAI）正在开发一种完全创新的模型，将保险欺诈/行为分析/侦探工作/机器学习/语言处理的独特团队专业知识与近实时呼叫“可信度”评分相结合，以识别欺诈性保险索赔。该公司计划将其软件作为现有通话录音平台的附加服务提供，从而提供：自动筛选每一个呼叫，将调查效率提高60倍/数据透明度/增强客户成果/保护易受攻击的客户/确保法规合规性/改进反欺诈措施。CAI处于有利地位，可以利用这一机会，已经建立了基本的线性算法，能够检测人类行为中的50个细微差别，以理解情境化情感，成功地使用手动转录的呼叫数据识别欺骗。作为一个新的合资企业，结合智能语音的现有和成熟的语言处理能力与斯特雷纽斯的行为分析专业知识，CAI将开发一个软件系统，能够识别人类互动使用GPU供电的自动语音识别成绩单。语音生物识别软件提供商已经表达了兴趣，他们提供开放访问其API的数据采集，并确认愿意试用该解决方案的联络中心。在Innovate UK的支持下，需要一个为期12个月的研发计划，以在封闭的现场环境中提供经过验证的系统，确定基于代理和呼叫者交互的呼叫者可信度评估的关键市场。CAI有可能破坏具有全球开发潜力的语音生物识别和反欺诈市场。该项目将为CAI带来出口导向型增长，可观的投资回报率（12700%），增加就业（增加25个工作岗位）和进一步的研发机会。"