Page 116 - AI for Good-Innovate for Impact Final Report 2024
P. 116

AI for Good-Innovate for Impact



                      (continued)

                       Domain         Business, economy
                       Model Training  •  Main training process: The proposed method utilizes a multi-interest and
                       and fine-tuning   entity-oriented pre-training architecture designed to learn generalized
                                         knowledge across various granularities. This architecture benefits from
                                         the inclusion of structural information in the entity graph, setting the
                                         stage for effective knowledge transfer.
                                      •  Main fine-tuning process: After the pre-training phase over source
                                         domains, prototype learning is integrated. This involves the use of a
                                         contrastive prototype learning module and a prototype enhanced
                                         attention mechanism. These components work together to improve the
                                         representations of users and items by leveraging adaptive knowledge
                                         utilization.
                       Case Studies   China (i.e. SDG 8,10).

                       Testbeds or    APE is deployed on Alipay Mini Program Cloud: link
                       pilot deploy-  APE is published in the publicly available paper PEACE: link
                       ments



                      25�2� Use case description


                      25�2�1  Description:

                      Amidst the surge in e-commerce, accelerated by the COVID-19 pandemic, SMEs are
                      increasingly turning to online platforms like mini-programs to maintain operations. However,
                      they encounter stiff competition from larger companies in reaching their desired customers.
                      Ant Personalization Engine (APE), an autonomous machine learning system, is thus proposed,
                      empowering small and medium-sized enterprises (SMEs) to connect with their target customers
                      more effectively, despite resource constraints.

                      APE aims to mitigate this inequality by providing SMEs with a personalized recommendation
                      engine that leverages large models, knowledge graph technology and AI-generated content
                      to enhance user engagement and drive business growth. APE delivers precise personalized
                      recommendations tailored specifically for SME-owned mini-programs. Additionally, integration
                      of AI-generated content enhances user experience, while multi-dimensional cold-start
                      technology ensures rapid engagement with new users, driving increased visibility and traffic.
                      Moreover, MLops automation streamlines operations, reducing manual intervention and
                      cutting operational costs.

                      The engine's personalized recommendation accuracy is currently being enhanced, while further
                      expansion across industry sectors and mini programs is in progress.

                      Across the world, SMEs play a vital role, contributing substantially to tax revenue, GDP,
                      technological innovation, and urban employment. They serve as the backbone of national
                      economic  and  social  development,  playing  a  crucial  role  in  expanding  employment
                      opportunities and improving living standards for people worldwide.

                      This case study is intricately linked to United Nations Sustainable Development Goal 8(Decent
                      Work and Economic Growth) and Goal 10(Reduced Inequalities). Goal 8, which focuses on
                      promoting sustained, inclusive, and sustainable economic growth, emphasizes the importance




                  100
   111   112   113   114   115   116   117   118   119   120   121