Evolving Telecoms scope 3 decarbonisation: an open-access emissions datasource powered by Vision Machine Learning

**问题：**英国电信业每年340万吨的二氧化碳排放量中，大部分都在“范围3”之内。解决这些间接排放问题相当于每年从英国道路上移除170万辆汽车。为了脱碳，电信运营商迫切需要可靠的供应链排放数据--但目前，现有的直接数据远远超过了人们可以轻易获取的数据。大量重要的供应商数据以非结构化的视觉格式隐藏在在线企业PDF报告中，现有的昂贵的人工智能方法不能优先准确地提取。因此，对于运营商和供应商来说，实现净零的工作并没有得到优化，也没有得到回报，因为会计仍然基于行业平均排放因素：一个巨大的拖累因素。**创新：**利用开创性的视觉机器学习来转换电信Scope-3数据：一个新的开放访问平台。通过开创性的基于视觉的机器学习技术，菌丝体从公司报告中定位并提取难以获取的“非结构化”排放数据，并促进无缝访问，以减少对行业估计的依赖，转变准确性，并帮助道德行为者竞争。目前，菌丝体与谢菲尔德大学合作，处于开发的第二阶段，菌丝体是第一个训练基于视觉的机器学习技术的平台，专门处理非结构化排放数据。它还将首先培训一个模块，专门识别与电信范围3中的脱碳努力相关的数据，并通过开放访问平台和面向大量用户的订阅应用编程接口(API)提供这些数据。作为由专有ML创新支持的专业模块，在提取电信运营商认真对待脱碳Scope 3所需的数据方面，菌丝体有可能超过像ChatGPT这样成本效益较低的服务的准确性。**数据访问民主化以加速净零**与专注于数据访问货币化的平台不同，菌丝体优先考虑开放访问和用户控制数据配置文件。之前数据被“埋藏”的公司被激励改善上市，增加自己的详细数据，并相应地进行竞争；承诺在排放报告方面提高质量和透明度，并挑战现有数据公司发展其服务。**改进产品**我们的机器学习模块已经建立，测试正在进行中。接下来，我们将训练模块识别电信范围3中的排放数据，并提取能够在这一棘手领域支持脱碳的非结构化数据。这些数据将通过我们即将到来的开放获取平台提供给任何需要它的人，我们将开发一个API，通过订阅向大量企业用户提供这些数据。

**The problem:**The majority of the UK telecoms sector's 3.4 million tons of CO2e annual emissions fall in 'Scope 3'. Addressing these indirect emissions would be comparable to removing 1.7 million cars from UK roads annually.To decarbonise, telecoms operators urgently need reliable data on supply chain emissions - but currently, far more direct data exists than can be readily accessed. Vast amounts of important supplier data are buried in unstructured visual format in online corporate PDF reports that existing, expensive AI methods don't prioritise extracting accurately. So for operators _and_ suppliers, work towards net zero isn't optimised _or_ rewarded because accounting is still based on industry average emissions factors: a huge drag factor.**The innovation:**Transforming Telecoms Scope-3 data with pioneering vision Machine Learning: a new open-access platform.Through pioneering vision-based Machine Learning techniques, Mycelium locates and extracts hard-to-access 'unstructured' emissions data from corporate reports and facilitates seamless access, to reduce reliance on industry estimates, transform accuracy and help ethical actors compete.Currently in its second phase of development in partnership with the University of Sheffield, Mycelium is the first platform to train vision-based machine learning techniques specifically to process unstructured emissions data. It will also be first to train a module specifically to recognize data relevant to decarbonisation efforts in Telecoms scope 3, and make it available via an open-access platform, and subscription Application Programming Interface (API) for large-volume users. A specialist module supported by proprietary ML innovations, Mycelium has the potential to surpass the accuracy of less cost-effective services like ChatGPT in extracting the data Telecoms operators need to get serious about decarbonising scope 3\.**Democratising data-access to accelerate Net Zero**Unlike platforms focused on monetising data access, Mycelium prioritises open access and user control of data profiles. Companies whose data was previously 'buried' are incentivised to improve their listing, add their own detailed data and compete accordingly; promising higher quality and transparency in emissions reporting, and challenging incumbent data companies to evolve their services.**Refining the offering**Our Machine Learning module is built, and testing is ongoing. Next we'll train the module to recognise emissions data in Telecoms scope 3 and extract the kind of unstructured data that can support decarbonisation in this tricky area. This data will be available to anyone who needs it via our forthcoming open-access platform, and we'll develop an API to make it available by subscription for large-volume corporate users.