SmartFi项目：为金融科技新闻探索AI_ML（英）

Project SmartFi Exploring AI/ML for FinTech NewsIN COLLABORATION WITH SYNTASA, POWERED BY GOOGLE CLOUDPowered byPublic Disclosure AuthorizedPublic Disclosure AuthorizedPublic Disclosure AuthorizedPublic Disclosure AuthorizedABSTRACTThe World Bank Finance and Technology Department, in collaboration with The World Bank Technology and Innovation Lab, partnered with Google Cloud and Syntasa Inc. to learn how artificial intelligence and machine learning could enhance the news sourcing and sentiment of FinTech topics globally. This outcome report shares the key learnings and insights as a part of the exploration and development of a prototype.ACKNOWLEDGEMENTSThe key learnings outlined in this report were prepared by the Project SmartFi (Smart Finance) team.World Bank Treasury Finance and Technology (TREFT): Paul Snaith, Patrick Cheng, Jaskaran SinghWorld Bank Technology and Innovation Lab (ITSTI:) Yusuf Karacaoglu, Stela Mocan, Mora Farhad, Mahesh Chandrahas Karajgi, Oleksandra Postavnicha, Yujuan SunWorld Bank Corporate Procurement: Sanjay Colaco, Shweta MesipamSyntasa Incorporated: Shawn Zargham, Michael Finn, Kyle Witt, James Wilson, Eric Bugin, Kareem Sharaf, Ted BlakeGoogle Cloud: Ryan Wright, Rajat GuptaContentsAbbreviations and Acronyms vSection 1: Overview 1Executive Summary 1Project Background 3Project Team & Sponsor 4Section 2: Exploration with Artificial Intelligence for Financial News 5Research Approach 5Business Challenge Scope 6Section 3: Collaboration with Google Cloud and Syntasa 9Rapid Prototyping with Technology Partners 9Solution Overview and Key Results 14Technical Approach (Syntasa) 22Section 4: Learning Outcomes and Future Considerations 37Technical Learnings for World Bank 37Business Learnings and Outcome 42Appendix A: Narrative Dashboard Features 46Appendix B: Reference Data 50Appendix C: Brandwatch 55Appendix D: SmartFi – Trusted Domains Technical Details 58Appendix E: SmartFi – Uncertain Domains Technical Details 62Appendix F: SmartFi – Chinese Language Technical Details 65FIGURES AND TABLESTable 2.1 6Figure 3.1: Syntasa Solution 10Figure 3.2: Modeled Mentions 16Figure 3.3: Word Cloud 17Figure 3.4: Domain Source 18Figure 3.5: Domain and PDF Sourcing 19Figure 3.6 : Trending Topics 20Figure 3.7: Sentiment Validation 21Figure 3.8: Sentiment Model Explainability 21Figure 3.9: Solution Architecture 23Figure 3.10: Data and AI pipeline 24Figure 3.11: Chinese Language App Configuration 25Figure 3.12: Topic Modeling Parameters 27Figure 3.13: Dashboard Trending Phrases 28Figure 3.14: Sentiment Explainability 29Figure 3.15: Sentiment Validation 30Figure 3.16: Language Translation Performance 32Figure 3.17: PDF Sourcing 33Figure 3.18: Sentiment Explainability 34Figure 3.19: Solution Architecture 35Figure 4.1: Topic Modeling 38Figure 4.2: Topic Modeling Explainer 39Table 4.1: Sentiment Analysis Models 40Project SmartFi: Exploring AI/ML for FinTech NewsivAbbreviations and AcronymsAbbreviationDescriptionAI Artificial IntelligenceAPI Applic