Rikoslaki SOTA
Rikoslaki SOTA is a source-grounded search, browsing and analysis system for Finnish legislation. Its first limited dataset is the Finnish Criminal Code.
What is it?
The project aims to build a structured and historical legal-data layer from Finlex material that can be searched deterministically and whose results can be traced back to the original legal sources.
The Criminal Code is the project's first limited dataset. The long-term goal is a historical search engine and local data mirror covering Finnish legislation more broadly.
What problem does it solve?
A generative language model is not, by itself, a reliable legal database. In legal use, it must be possible to determine which law, provision and point in time an answer is based on. Legislation also changes over time.
Rikoslaki SOTA is designed to separate the source material, retrieval process and AI-generated interpretation. The legislation itself forms the system's primary information layer. AI can operate on top of that layer, but it does not replace the original legal source.
How it works
Finlex material → ingestion and normalisation → historical legal-data mirror → search → source references → AI-assisted interpretation
Legislative material obtained from Finlex is ingested into a structured database. Version and validity information is handled so that search results can be associated with the correct legal provision and the correct point in time. AI-assisted functions can then be built on top of this retrieval layer using the system's own source material.
Who is it for?
- legal professionals and lawyers
- researchers and educational institutions working with legislation
- publishers and public authorities requiring precise source references
- developers building legal applications
Technical implementation
The architecture emphasises deterministic retrieval, temporal correctness and verifiable source references. PostgreSQL forms the foundation of the structured legal-data layer, and the retrieval system is designed to remain source-grounded.
Lexical retrieval and result fusion may later be complemented by semantic search methods. AI-based methods operate as additional layers, but they are not allowed to replace the original legal source.
Current status
Rikoslaki SOTA is in research and development. The Criminal Code browser, ingestion pipeline and retrieval architecture have already been developed and tested, but the system is not marketed as a finished legal advisory service.
Next step
The next step is to expand the Finlex data mirror, strengthen the ingestion process and temporal version management, and continue measuring retrieval quality. The architecture is also being developed so that additional areas of Finnish legislation can later be added alongside the Criminal Code.
From the Criminal Code to Finnish legislation
The Criminal Code is the project's first limited legal corpus. The long-term objective is to build a historical search engine for Finnish legislation using Finlex data, allowing users to search both current legislation and legislation that was in force at earlier points in time.
This data layer could later support a source-grounded legal assistant and an API through which legal data and retrieval functions could be made available to other applications.
Commercialisation possibilities
The project is investigating the possibility of productising an API service on top of the legal-data layer. This is a potential future commercial direction, not a currently released service.
- source-grounded legal assistant
- historical legislation search service
- legal-data search API
- integrations with other legal services