Apparatus and method for matching POI entities
An apparatus is provided for matching a first point of interest (POI) dataset including first POI entities with a second POI dataset including a plurality of second POI entities in a 2D region. Each of the first and second POI entities includes coordinate data and one or more label tokens. The apparatus includes processing circuitry configured to determine, for each first POI entity, a subset of the plurality of second POI entities. The processing circuitry is configured to determine, for each first POI entity, a composite matching score with one or more second POI entities. The composite matching score is based on a coordinate matching score and a label matching score, the coordinate matching score depending on a geographical distance and the label matching score depending on a local context relevance. The processing circuitry is configured to match each first POI entity with the second POI entities of the subset.
1. An apparatus for matching a first point of interest (POI) dataset including one or more first POI entities with a second POI dataset including a plurality of second POI entities in a 2D region, each respective first POI entity of the first POI entities comprising first coordinate data and one or more first label tokens indicative of a label of the respective first POI entity, each respective second POI entity of the second POI entities comprising second coordinate data and one or more second label tokens indicative of a label of the second POI entity, the apparatus comprising:
processing circuitry configured to:
determine, for each respective first POI entity of the one or more first POI entities based on the first coordinate data of the one or more first POI entities, a subset of the plurality of second POI entities;
determine, for each respective first POI entity of the one or more first POI entities, a composite matching score with each of one or more respective second POI entities of the subset of the plurality of second POI entities, wherein the composite matching score is based on a coordinate matching score and a label matching score, wherein the coordinate matching score depends on a geographical distance between the respective first POI entity and the respective second POI entity, and wherein the label matching score depends on a local context relevance of the one or more first label tokens of the respective first POI entity; and
match, based on the composite matching score, each respective first POI entity of the one or more first POI entities with the second POI entities of the subset of the plurality of second POI entities.
2. The apparatus of claim 1 , wherein each of the one or more first POI entities further comprises one or more first address tokens and each of the plurality of second POI entities further comprises one or more second address tokens,
wherein the composite matching score determined for each respective first POI entity of the one or more first POI entities is further based on an address matching score, wherein the address matching score depends on a local context relevance of the one or more first address tokens of the respective first POI entity.
3. The apparatus of claim 2 , wherein the processing circuitry is further configured to:
generate, for each subset of the plurality of second POI entities, a document based on the one or more second label tokens and/or the one or more second address tokens of the second POI entities belonging to the subset of the plurality of second POI entities, and
determine the local context relevance of the one or more first label tokens and/or the local context relevance of the one or more first address tokens of the respective first POI entity as an inverse document frequency (IDF) of the one or more first label tokens and/or the one or more first address tokens of the respective first POI entity in the respective document.
4. The apparatus of claim 3 , wherein the processing circuitry is configured to determine the local context relevance of the one or more first label tokens and/or the local context relevance of the one or more first address tokens of the respective first POI entity based on a local IDF-weighted Jaccard similarity.
5. The apparatus of claim 2 , wherein the processing circuitry is configured to determine the composite matching score as a weighted sum of at least two of: the coordinate matching score, the label matching score, or the address matching score.
6. The apparatus of claim 1 , wherein the one or more first POI entities comprises a first POI entity, and wherein the processing circuitry is configured to determine, based on the first coordinate data of the first POI entity, the subset of the plurality of second POI entities by determining the second POI entities of the plurality of second POI entities located in a vicinity of the first POI entity.
7. The apparatus of claim 1 , wherein the processing circuitry is configured to determine, for each of the one or more first POI entities of the first POI dataset based on the first coordinate data of the one or more first POI entities, the subset of the plurality of second POI entities of the second POI dataset by spatially partitioning the region into a plurality of partitioning cells.
8. The apparatus of claim 7 , wherein, for each partitioning cell, a product of the number of first POI entities and second POI entities located in the partitioning cell is smaller than or equal to a predefined threshold value.
9. The apparatus of claim 7 , wherein the spatial partitioning is a quadtree partitioning of the region.
10. The apparatus of claim 9 , wherein each partitioning cell has a rectangular shape and comprises a frame-shaped boundary region surrounding the partitioning cell, wherein the ratio of the area of the boundary region to the area of the partitioning cell depends on a local density of the plurality of first POI entities and/or the plurality of second POI entities.
11. The apparatus of claim 1 , wherein the label matching score is further based on a label embedding score between a first label embedding vector based on the one or more first label tokens of the first POI entity and a second label embedding vector based on the one or more second label tokens of the second POI entity.
12. The apparatus of claim 11 , wherein the processing circuitry is further configured to reduce the number of second POI entities of each subset of the plurality of second POI entities by determining the label embedding score between each of the one or more first POI entities and the one or more second POI entities of the subset of the plurality of second POI entities and removing at least one of the one or more second POI entities from the subset that has a lowest label embedding score.
13. The apparatus of claim 1 , wherein each of the first POI entities further comprises one or more first types and each of the second POI entities further comprises one or more second types, and wherein the composite matching score is further based on a type matching score between the one or more first types of the first POI entity and the one or more second types of the second POI entity.
14. The apparatus of claim 13 , wherein the type matching score is based on a type embedding score indicative of a similarity between a first type embedding vector based on the one or more first types of the first POI entity and a second type embedding vector based on the one or more second types of the second POI entity.
15. The apparatus of claim 14 , wherein the processing circuitry is further configured to reduce the number of second POI entities of each subset of the plurality of second POI entities by determining the type embedding score between each of the one or more first POI entities and the one or more second POI entities of the subset of the plurality of second POI entities and removing at least one of the one or more second POI entities from the subset that has a lowest type embedding score.
16. The apparatus of claim 14 , wherein the apparatus is further configured to implement an unsupervised machine learning scheme configured to determine the first type embedding vector based on the one or more first types of the first POI entity and the second type embedding vector based on the one or more second types of the second POI entity.
17. The apparatus of claim 1 , wherein the apparatus further comprises a memory configured to store the second POI dataset including the plurality of second POI entities.
18. The apparatus of claim 1 , wherein the apparatus further comprises a communication interface configured to receive the first POI dataset including the one or more first POI entities.
19. A method for matching a first point of interest (POI) dataset including one or more first POI entities with a second POI dataset including a plurality of second POI entities in a 2D region, each of the first POI entities comprising first coordinate data and one or more first label tokens, each of the second POI entities comprising second coordinate data and one or more second label tokens, the method comprising:
determining, for each respective first POI entity of the one or more first POI entities based on the first coordinate data of the one or more first POI entities, a subset of the plurality of second POI entities;
determining, for each respective first POI entity of the one or more first POI entities, a composite matching score with one or more second POI entities of the subset of the plurality of second POI entities, wherein the composite matching score is based on a coordinate matching score and a label matching score, wherein the coordinate matching score depends on a geographical distance between the respective first POI entity and the respective second POI entity, and wherein the label matching score depends on a local context relevance of the one or more label tokens of the respective first POI entity; and
matching, based on the composite matching score, each respective first POI entity of the one or more first POI entities with the second POI entities of the subset of the plurality of second POI entities.
20. The method of claim 19 , wherein each of the first POI entities further comprises one or more first address tokens and each of the second POI entities further comprises one or more second address tokens, and
wherein the composite matching score is further based on an address matching score, wherein the address matching score depends on a local context relevance of the one or more first address tokens of the respective first POI entity.