Novel research into language understanding to automate customs and freight paperwork

摩根大通2017年贸易展望报告估计，由于人员效率低下和缺乏审查，与贸易有关的财富500强公司每年产生超过810亿美元的不必要供应链成本。自中世纪以来，货运代理公司一直代表a)托运人组织运输，通过空运、海运或陆运，以及b)承运人将实物货物从A运往B。传统的劳动密集型纸质运输方式导致周转时间长，几乎没有正式的知识获取和高运营风险。麦肯锡咨询公司报告称，“在全球供应链中，电话和电子邮件仍然是主要渠道，就像几十年前一样”，CargoLogik指出，“目前的流程是巨大的时间和开销杀手……仍然主要是人工……这导致货运代理损失时间，客户体验笨拙、糟糕。”摩根大通估计，每笔发货通常涉及27个高度分散的文档实体。该项目旨在开发一个基于专家自然语言处理(NLP)的工具，从电子邮件中发现的非结构化文本中识别和提取关键信息，该工具将针对物流部门特定的语言和电子邮件通信进行培训和配置。在这样做的过程中，我们将开发能够创建结构化规则的技术，用于对照贸易文书对电子邮件中的自由格式文本进行交叉检查，并推动对全球贸易文书中的内容采取行动。

JP Morgan 2017 trade outlook report estimates that trade-involved Fortune 500 companies incur more than $81 billion of unnecessary supply chain costs each year due to human inefficiencies and lack of scrutiny. Since medieval times, freight forwarders have organized transportation on behalf of a) shippers, by air, ocean or land, and b) carriers, to get physical goods from A to B. The traditional labour- intensive paper driven approaches result in long turn-around times, little formal knowledge capture and high operational risk.McKinsey consulting report that "in global supply chains, telephones and email are still the dominant channels, just as they were decades ago", and CargoLogik note the "current processes are an immense time and overhead killer... still largely manual ...which result in lost time for freight forwarders and a clunky, poor customer experience." JPMorgan estimates that there are typically 27 highly fragmented document entities involved for every shipment made.This project aims to develop an expert natural language processing (NLP) based tool, to identify and extract key information from unstructured text found in emails, which will be trained and configured for the logistics sector specific language and email communications. In doing so we will develop technology capable of creating structured rules for cross-checking free-form text in emails against trade paperwork and driving action on the content found within global trade paperwork.