践美好，见未来：2021亚马逊云科技线上黑客松落幕
Practice the Good, See the Future: 2021 Amazon Web Services Online Hackathon Concludes
近日，2021 亚马逊云科技线上黑客松圆满结束。
Recently, the 2021 Amazon Web Services Online Hackathon came to a successful conclusion.
AI 应用的加速落地，智能驱动的科技创新带来了生活方式的变革与重塑，让美好生活变得有形和鲜活，这场线上举行的黑客松又有哪些值得关注之处呢？
With the accelerated implementation of AI applications and technology-driven innovation bringing transformation and reshaping to lifestyles, making a better life tangible and vibrant, what noteworthy aspects did this online hackathon have?
自今年 3 月开启以来，本次大赛以「AI 开发 · 践美好 · 见未来」为主题，围绕「AI + 品质生活」、「AI + 自动驾驶」、「AI + 体育运动」、「AI + 电竞游戏」四大赛题，吸引了多位高校学生、创业者、个人开发者、企业开发者报名，让开发创造无限可能。此外，大赛期间还设置了互动直播，邀请了亚马逊云科技资深技术专家莅临直播间，与开发者们进行了充分的互动交流。
Since its launch in March this year, the competition, themed "AI Development · Practice the Good · See the Future," revolved around four major topics: "AI + Quality Life," "AI + Autonomous Driving," "AI + Sports," and "AI + Esports." It attracted numerous university students, entrepreneurs, individual developers, and enterprise developers, enabling unlimited possibilities for development. Additionally, interactive live streams were set up during the competition, inviting senior technical experts from Amazon Web Services to the studio for full interaction and exchange with developers.
来自全球各大企业和高校的编程爱好者围绕「善用 AI，创造美好生活」、「以不变的 AI 核心，应对万变的自动驾驶场景」、「AI 为体育行业带来的新元素、新思想和新玩法」和「AI 加持，开启电竞加速度」四大赛题，结合至少一项亚马逊云科技产品服务，提交了多个优秀作品。
Programming enthusiasts from major global enterprises and universities submitted outstanding works based on the four themes: "Making Good Use of AI to Create a Better Life," "Responding to Ever-Changing Autonomous Driving Scenarios with Unchanging AI Core," "New Elements, New Ideas, and New Playstyles that AI Brings to the Sports Industry," and "AI Empowerment, Accelerating Esports." Each submission incorporated at least one Amazon Web Services product or service.
本次大赛共收到了 12 组选手的提交方案，经历了为期半年的比赛，4 支队伍突破了严格的筛选，获得了决赛入场券。
A total of 12 team proposals were received for this competition. After six months of competition, four teams passed strict screening and secured spots in the finals.
三强团队
The Top Three Teams
在最后的环节，黑客松的专家评委从赛题完成度、创意度、使用体验、商业价值、展示度、影响力、技术架构多个维度对选手们的作品进行综合考量，评选出了 3 支优胜队伍。
In the final round, expert judges from the hackathon comprehensively evaluated the works from multiple dimensions, including completeness of the problem-solving approach, creativity, user experience, commercial value, presentation, influence, and technical architecture, ultimately selecting three winning teams.
评审团队阵容
Judging Panel
7 月 6 日，主办方正式公布了决赛前三名获奖团队，获奖方案分别为「AI 会议便签条」、「减肥辅助机器人」和「多语言交流学习便捷助手」。
On July 6, the organizers officially announced the top three winning teams. The award-winning solutions were "AI Meeting Sticky Notes," "Weight Loss Assistant Robot," and "Multilingual Communication Learning Convenient Assistant."
获奖方案介绍
Introduction to the Winning Solutions
在工作和生活中，会议记录的分发和整理是一件费力不讨好的事情，对于想要了解具体开会内容的人，阅读连篇累牍的会议记录很难快速获取自身感兴趣的摘要内容。
In work and daily life, distributing and organizing meeting minutes is a thankless task. For those who want to understand specific meeting content, reading lengthy meeting records makes it difficult to quickly obtain summaries of topics they are interested in.
冠军团队「AI 会议便签条」利用公开的美国加州圣何塞市 2020 年市政会议记录（共计 45 份会议记录，842 页，1648 个段落，11 万字）作为训练数据集。通常政府的市政会议内容会以 PDF 的格式进行线上公开，人们很难根据具体内容进行检索和查阅。
The champion team, "AI Meeting Sticky Notes," used publicly available 2020 municipal meeting records from San Jose, California, USA (totaling 45 meeting records, 842 pages, 1,648 paragraphs, and 110,000 words) as their training dataset. Typically, government municipal meeting content is made publicly available online in PDF format, making it difficult for people to search and retrieve specific content.
团队对会议内容进行结构细分和文本预处理，把文本数据转换成便于机器学习算法直接使用的实值向量，并用自然语言处理的算法进行主题模型(Topic Modeling)，加入词嵌入（Word2Vec）技术进行分类特征处理，从而让公众可以对市政会议里感兴趣的主题进行跨时间线的检索与订阅，像翻阅便签条一样，无须阅读完整的会议记录而获取感兴趣的细分内容。作品未来的愿景是推广到企业内部，对于各部门会议的具体内容进行片断式的高效分类、检索和订阅，从而节省部分与会者的参与时间，以建立自动化的 AI 系统来提高开会和沟通的效率。
The team performed structural segmentation and text preprocessing on the meeting content, converting text data into real-valued vectors suitable for direct use by machine learning algorithms. They applied natural language processing algorithms for topic modeling and incorporated Word2Vec technology for classification feature processing, enabling the public to search and subscribe to topics of interest across timelines—much like flipping through sticky notes—without having to read complete meeting records to obtain segmented content of interest. The future vision for this work is to promote it within enterprises, enabling efficient fragment-based classification, retrieval, and subscription of department-specific meeting content, thereby saving some participants' time and establishing automated AI systems to improve meeting and communication efficiency.
在冠军方案的打造过程中，团队使用了多项亚马逊云服务，最终在一个较短的时间内建立了完善的方案，实现了理想的模型效果。团队特别提到，亚马逊云科技提供了极为全面的云服务解决方案，各种服务器、开发管理和云计算工具，应有尽有。比如 Amazon S3 / Amazon Simple Storage Service / Amazon Elastic File System 可用于储存用户上传文本数据，并添加到原始数据集，重新训练并优化文本搜索算法；亚马逊云科技 Textract 用于提取 PDF 原始数据，对手写文本可以自动识别，并且附带了许多数据分析信息，非常方便；而 IAM / 亚马逊云科技 Key Management Service 可根据团队成员的工种设置不同的应用权限，安全可靠且互不影响，并可共享同一个 Billing 账户，极大地方便了开发管理。此外，AmazonCloudWatch / Amazon Simple Notification Service 用于监控网站异常情况，便于相关人员及时维护，设置好监控指标以后，便可发送邮件或短信自动通知。
In developing the champion solution, the team utilized multiple Amazon Web Services, ultimately establishing a comprehensive solution in a relatively short time and achieving ideal model performance. The team particularly noted that Amazon Web Services provides an extremely comprehensive range of cloud service solutions, covering various servers, development management tools, and cloud computing tools. For example, Amazon S3 / Amazon Simple Storage Service / Amazon Elastic File System can be used to store user-uploaded text data and add it to the original dataset for retraining and optimizing text search algorithms; Amazon Textract is used to extract raw PDF data, automatically recognizing handwritten text and providing rich data analysis information, making it very convenient; and IAM / Amazon Key Management Service allows setting different application permissions based on team members' roles, ensuring security and independence while sharing a single billing account, greatly facilitating development management. Additionally, Amazon CloudWatch / Amazon Simple Notification Service monitors website anomalies for timely maintenance by relevant personnel, automatically sending email or SMS notifications once monitoring metrics are configured.
本次冠军团队的奖品内容包含：3000 美金亚马逊云科技服务抵扣券、4999RMB 九号(Ninebot)One Z6 单轮平衡车 1 台、3598RMB GoPro HERO9Black 套组 1 套等。
The champion team's prizes included: a $3,000 AWS service credit, one 4999 RMB Ninebot One Z6 unicycle, one 3598 RMB GoPro HERO9 Black set, etc.
亚军团队「减肥辅助机器人」的产品功能主要有两个：一是信息问询，输入食物或者运动，输出食物对应的热量，以及运动对应会消耗的热量；二是减肥咨询，可通过用户输入身高体重，计算出静息代谢，也可以通过用户输入目标体重和现在的体重，计算出每天应该消耗多少热量，摄入多少热量。
The runner-up team, "Weight Loss Assistant Robot," has two main product functions: first, information inquiry—inputting food or exercise to output the corresponding calories of the food and the calories consumed by the exercise; second, weight loss consultation—calculating resting metabolism based on user-input height and weight, and calculating daily calorie consumption and intake targets based on current and target weights.
这款产品在打造过程中使用了亚马逊的机器学习云服务 Amazon translate，团队成员介绍说：「这些训练数据都来自于自己减肥时积累的数据。」产品采用了最流行的微信公众号作为产品入口，并在内部分为 NLP 模块、减肥信息结果计算模块与 Amazon translate 几个模块，结构清晰。通过使用亚马逊云科技服务，团队可以方便地直接调用很多成熟的能力，减少开发成本，提升开发速度，对于小型产品的发展非常有用。
In developing this product, the team used Amazon's machine learning cloud service Amazon Translate. Team members said, "All this training data came from the data we accumulated while losing weight ourselves." The product adopted the popular WeChat official account as its entry point, internally divided into an NLP module, a weight loss information calculation module, and an Amazon Translate module, with a clear structure. By using AWS services, the team could conveniently directly invoke many mature capabilities, reducing development costs and accelerating development speed, which is very useful for the growth of small-scale products.
亚军团队的奖品内容包含：2000 美金亚马逊云科技服务抵扣券、3598RMB GoPro HERO9Black 套组 1 套、2499RMB 索尼 (SONY) 蓝牙降噪游戏耳机 1 只等。
The runner-up team's prizes included: a $2,000 AWS service credit, one 3598 RMB GoPro HERO9 Black set, one 2499 RMB Sony Bluetooth noise-canceling gaming headset, etc.
季军团队「多语言交流学习便捷助手」的设计初衷来源于人们去其他国家旅游或者出差时对于异国语言交流的需求，作品主要实现多种语言的输入及对应的翻译输出。考虑到实际使用时的便捷形式，这款产品除了支持文本输入以外，还支持语音输入。另外，本产品也可作为语言学习工具，让用户能够比较方便地学习和锻炼简单语句的翻译。
The third-place team, "Multilingual Communication Learning Convenient Assistant," was inspired by the need for cross-language communication when people travel or go on business trips to other countries. The product primarily supports input in multiple languages with corresponding translated output. Considering convenience in practical use, besides text input, the product also supports voice input. Additionally, this product can serve as a language learning tool, allowing users to conveniently learn and practice translating simple sentences.
在制定方案时，团队使用了 Amazon S3、Amazon Transcribe、Amazon Translate 三项亚马逊云提供的机器学习服务，Amazon Transcribe 负责将存储在 Amazon S3 中的音频文件转换为准确、完整的标点符号文本，不需要开发人员熟悉自然语言处理的相关理论，加快了系统开发的效率，而 Amazon Translate 帮助产品实现了高准确率的文本翻译。产品主要分为前端交互界面和后端逻辑处理部分：交互界面负责接收用户的输入信息，当有输入信息时，触发后端服务进行对应的逻辑处理；后端逻辑处理负责根据输入信息，使用 Amazon 云科技工具进行处理，得到输出结果返回给前端的接收接口。在使用语音输入时，将 Amazon S3 作为数据存取的中转站，设定 bucket 存放和取出语音信息及转换结果。
In formulating the solution, the team used three AWS machine learning services: Amazon S3, Amazon Transcribe, and Amazon Translate. Amazon Transcribe is responsible for converting audio files stored in Amazon S3 into accurate, fully punctuated text, without requiring developers to be familiar with NLP theories, thus accelerating system development efficiency, while Amazon Translate helps the product achieve highly accurate text translation. The product is mainly divided into a front-end interactive interface and a back-end logic processing module: the interactive interface receives user input, and when input is received, it triggers the back-end service to perform corresponding logic processing; the back-end logic processing is responsible for using AWS tools to process the input and return the output results to the front-end receiving interface. When using voice input, Amazon S3 serves as a data transfer station, with buckets configured to store and retrieve voice data and conversion results.
季军团队的奖品内容包含：1000 美金亚马逊云科技服务抵扣券、2499RMB 索尼(SONY) 蓝牙降噪游戏耳机 1 只等。
The third-place team's prizes included: a $1,000 AWS service credit, one 2499 RMB Sony Bluetooth noise-canceling gaming headset, etc.
此外，所有参赛团队均获得了主办方提供的「阳光普照奖」，奖品包括 300 美金的亚马逊云科技服务抵扣券等。
In addition, all participating teams received the "Sunshine Award" provided by the organizers, which included prizes such as a $300 AWS service credit.
小到一「码」在手，出行无忧，大到万物互联，智能升级，人工智能已经成为人类生活中的技术底色。在这次圆满完结的 2021 亚马逊云科技线上黑客松中，我们见证了「善用 AI」提升生活品质的应用，也有走向精细化的自动驾驶应用场景探索，以及智能加持的技术创新与文娱、体育场景。未来，秉承着践美好、见未来的初心，亚马逊云科技将鼓励越来越多的开发者参与到赛事之中，用技术创新推动社会的进步。
From having a single code to travel worry-free, to the Internet of Everything and intelligent upgrades, artificial intelligence has become the technological foundation of human life. In this successfully concluded 2021 Amazon Web Services Online Hackathon, we witnessed applications that "make good use of AI" to improve quality of life, explorations of autonomous driving scenarios moving toward refinement, and technological innovations empowered by intelligence in entertainment and sports scenarios. In the future, staying true to the original intention of practicing the good and seeing the future, Amazon Web Services will encourage more and more developers to participate in such events, using technological innovation to drive social progress.
