Hi everyone!
This is a blog post providing an update on the progress of my project, which aims to improve the search experience in Japan. My project is 12 weeks long, and we are currently in the 7th week.
How did we approach it?
Based on our research, we identified two key areas for enhancing the search experience. These two aspects are interconnected, and we are working on resolving them simultaneously.
Searching for addresses: Currently, Nominatim focuses on the addr:street and addr:place tags when searching for addresses. However, in Japan, addresses are primarily based on block addresses, and the street component is less significant. Therefore, we need to ensure that Nominatim can appropriately assign the correct parent when conducting searches.
Importing data: Nominatim generates a database from an OSM planet file. To accommodate the block address system in Japan, we are adding a new sanitization function to adjust Japanese addresses. This will ensure that the data generated from the OSM planet file aligns with the block address structure, similar to the changes made on the searching side.
Progress Update
Through modifications that enable the segmentation of Japanese search terms and mapping them to block addresses, we have successfully enabled address searches. It is possible to search for such addresses (神奈川県幸区南加瀬3丁目 8-33). Also, as demonstrated in Fig. 3 and Fig. 4, we were able to associate addresses with the correct block, whereas previously, streets were prioritized. I am going to submit a pull request to incorporate the code improvements I made into the Nominatim codebase.
This image shows the state of the code before modification (Fig. 1).
This image shows the state of the code after modification (Fig. 2).
This image shows the enlarged view of Fig. 1 (Fig.3).
This image shows the enlarged view of Fig. 2 (Fig.4).
Next Steps
Firstly, we will measure the accuracy and speed improvements resulting from the segmentation of search terms. This will help us assess the actual impact of these changes. Additionally, I also plan to document the process using Japanese addresses as an example, enabling the customization of these improvements for other countries as well.
I would like to thank my mentors, Sarah Hoffman (@lonvia) and Marc Tobias (@mtmail), for their invaluable guidance throughout the implementation of this project. Please feel free to provide any feedback or suggestions. Thank you for reading!