The end of last year saw the release of the 3.6.0 version of Nominatim. It concludes last years effort to improve the address information that is returned to you when you search for something. We have gradually changed how Nominatim organises address information and how it interprets information on OSM place nodes and administrative boundaries. The new system also takes into account that today the existence of administrative boundaries is rather the norm in OSM than the exception. When the original code was written years ago, this was still different. There is still some work to do to look through the data of each country and adapt the ranking system now in place to the local customs but these are more smaller changes we can do gradually.

So where to go from here? I’m happy to announce that the NGI Zero Discovery fund has accepted a proposal to partially support the development of Nominatim over the coming year. The main focus of this work will be on adding language-awareness for query parsing and place name interpretation. The goal is to improve the quality of search results across all languages but especially for languages that use letters beyond a simple A-Z. We want to get to an architecture where users can more easily contribute their local knowledge to improve searching in their language or region. And it is worth mentioning that reworking the query parsing is also an important step towards the much demanded fuzzy search.

Here is a more detailed list of steps to be tackled in this project:

  • Introduce standard libraries for normalization. Normalization is the process of removing features of a query or name string that are not relevant for search. We plan to replace the very simple hard-coded character lookup table with a more flexible and standardized approach using libicu.
  • Configurable abbreviation handling. Abbreviations (like ‘st’ for street) are very common in geographic names. Nominatim has a relatively short hard-coded list of abbreviations which is applied with little regard for context. We want to make abbreviations configurable for an installation and allow to add annotations (like language and grammatical structure).
  • Develop a show-case language-sensitive parsing algorithm Finally we want to put everything together and creating a language-sensitive parsing component for Nominatim. The goal is to use this as the new default parsing algorithm but also for it to be usable as an example for users who want to create their own extensions.

In addition, there are some secondary goals for the project:

We finally want to make Nominatim packagable for distributions. You shouldn’t need to have the source tree lying around on your production machine. Setting up the software should become as simple as apt install nominatim. We want to get away from the monolithic, hard-coded architecture and make it easier to extend Nominatim and customize it for your particular use case. And while the documentation is slowly growing on nominatim.org there is always room for improvement to make it more detailed and accessible for everybody.

There is a lot of work to be done and you can help to speed up development by becoming a supporter yourself. Have a looked at the funding page to learn about the different options to support Nominatim.