面向众源POI融合的溯源建模与更新方法研究

The construction of place-name databases is the basis for the move of geographic location representation from geometric reference systems to linguistic descriptions. The crowdsourced point of interest (POI) on the Web is an important data source for constructing and updating place-name databases, and constitutes the core basic information for intelligent location-based services. In the crowdsourcing environment, POI comes from diverse sources and varies largely in data quality. Thus how to effectively record the provenance information of multi-source POI is of great significance for judging whether these POI is reliable and optimizing the process of POI conflation and updates. This project proposes an interoperable information model for expressing and processing crowdsourced POI provenance information, in order to address provenance tracking challenges resulted from POI heterogeneities in place names, taxonomies, and versions. With the interoperable model, provenance information will be automatically captured and tracked during the POI conflation process. After that, methods of optimizing and updating conflated POI database based on the analysis of captured provenance information will be studied, in order to make the database reliable and keep it up-to-date. The research will contribute to current POI conflation method with standardized provenance semantics, and build a backtracking mechanism on POI locations, sources and evolution processes. The results will support location-based social sensing in the context of big data, and promote the applications of intelligent location-based services.

地名实体库是地理位置从几何表达走向语义与自然语言表达的基础。互联网众源POI（兴趣点）是地名实体库构建与更新的重要来源，并成为面向大众位置智能服务的核心基础信息。在众源环境下，POI来源多样，质量各异，如何有效记录POI的溯源信息，对于判断POI是否可信与可靠、优化POI融合与更新流程具有重要意义。本项目从互操作的理念出发，设计适应于不同来源POI溯源关系表达与处理的共享信息模型，突破POI演变历史中命名方式差异、分类体系差异、时间版本差异等溯源记录难题，建立支持来源回溯的众源POI融合实体库，在此基础上，开展基于溯源的POI融合优化与更新方法研究，实现可信众源POI实体库的快速更新。研究成果将为众源POI融合实体库提供规范化的溯源语义表达，构筑对众源POI历史位置、来源及演变等信息的回溯机制，推动众源地理信息在智能位置服务领域的深入应用，为大数据背景下基于位置语义的社会感知提供基底支持。

众源POI作为地名实体库的重要来源，是面向大众位置智能服务的核心基础信息。融合多源POI有助于实现信息互补并保持现势性。但在众源环境下，不同来源的POI质量各异，且存在命名方式差异、分类体系差异、时间版本差异等问题。为此，本项目从数据溯源视角出发，设计了适应于不同来源POI溯源关系表达与处理的共享信息模型，提出了多源异构POI匹配方法，建立了支持来源回溯的众源POI位置知识库。在本项目的资助下，课题组共发表学术论文8篇，申请发明专利5项。依托本项目的部分研究成果，负责人获得2022年中国测绘科技进步特等奖(排名5)。负责人担任国际著名SCI刊物Remote Sensing客座编辑。研究成果为众源POI融合的位置知识库库提供规范化的溯源语义表达，构筑了对众源POI历史位置、来源及演变等信息的回溯机制，有望推动众源地理信息在智能位置服务领域的深入应用，为大数据背景下基于位置语义的社会感知提供基底支持。