An End-to-End Point of Interest (POI) Conflation Framework

Abstract

Point of interest (POI) data serves as a valuable source of semantic information for places of interest and has many geospatial applications in real estate, transportation, and urban planning. With the availability of different data sources, POI conflation serves as a valuable technique for enriching data quality and coverage by merging the POI data from multiple sources. This study proposes a novel end-to-end POI conflation framework consisting of six steps, starting with data procurement, schema standardisation, taxonomy mapping, POI matching, POI unification, and data verification. The feasibility of the proposed framework was demonstrated in a case study conducted in the eastern region of Singapore, where the POI data from five data sources was conflated to form a unified POI dataset. Based on the evaluation conducted, the resulting unified dataset was found to be more comprehensive and complete than any of the five POI data sources alone. Furthermore, the proposed approach for identifying POI matches between different data sources outperformed all baseline approaches with a matching accuracy of 97.6% with an average run time below 3 min when matching over 12,000 POIs to result in 8699 unique POIs, thereby demonstrating the framework’s scalability for large scale implementation in dense urban contexts.

Conflating point of interest (POI) data: A systematic review of matching methods

“…These preprocessing steps generally help remove string differences due to syntactic issues, such as different uses of upper and lower cases or different word orders, and help the similarity measurement focus on the remaining and more meaningful parts of the POI names. Novack et al (2018) and Low et al (2021) used token sort ratio which tokenizes POI names into individual words, sorts them in alphabetical order to form new strings, and then calculates the Levenshtein distance between the new strings. Similar methods, such as token set ratio (based on common tokens in two POI names while ignoring token orders), were also used in Piech et al (2020).…”

“…From time to time, researchers may want to merge two or more POI datasets in order to obtain a better representation of the places in their study areas. This process is generally called POI conflation (Low et al, 2021;McKenzie et al, 2014). A main reason for POI conflation is that different POI datasets may have different attribute focuses, place coverages, and data quality.…”

“…Many POI conflation methods have been developed in previous research (Lamprianidis et al, 2014;Low et al, 2021;Piech et al, 2020;Yang et al, 2014). However, there lacks a systematic review and it is difficult for researchers new to POI conflation to quickly grasp and use these methods.…”

A System for Aligning Geographical Entities from Large Heterogeneous Sources

“…Low et al [14] proposed a complete POIs conflation framework, from data catering to data verification. They also used POIs' attributes information to match POIs.…”

“…This is performed under the assumption that types' taxonomies from different data sources are consistent and not noisy. In addition, although Low et al [14] claimed large-scale application of their conflation framework, the maximum scale of the used dataset does not exceed 12,000 POIs. The scalability limitation is mainly due to the manual intervention of human experts in the verification step.…”