A VALUED CUSTOMER OF THIRDEYE DATA

European Court of Human Rights: making case law searchable by meaning

The European Court of Human Rights was established in 1959 to oversee enforcement of the Convention for the Protection of Human Rights and Fundamental Freedoms. Its case law is published through HUDOC, which is used by lawyers, researchers, journalists and governments across Europe. A body of case law is only as useful as the ability to find the relevant part of it.

LOCATION

Strasbourg, France

COMPANY SIZE

International Court

INDUSTRY

Government and Public Sector

From keyword search to semantic search

HUDOC had a keyword search. Keyword search works when the person searching already knows the vocabulary used in the document they are looking for. In law that assumption fails constantly, because the same principle is expressed differently across decades, across languages and across judicial traditions, and someone researching a concept rarely knows which words a particular judgment chose.

ThirdEye Data built the semantic layer that sits over the existing search. Rather than matching words, it groups the case law by what the judgments are actually about, and connects the legal concepts the Court works with to the language that appears in the text. That lets someone searching for a concept find the judgments that deal with it, whatever words those judgments happen to use.

On top of that the system produces, for each document, a summary, the entities specific to this material, the key phrases, and a generated title, so that a result list tells a researcher enough to judge relevance without opening every document.

We also built the feedback loop. Users can rate what comes back, and concepts can be added or removed as the Court’s own understanding of its material develops. A legal corpus is not static and a fixed model would have aged badly.

A second phase, from April 2020, took the work live on the HUDOC site.

Case law is searchable by legal concept rather than by matching words.

This engagement came to ThirdEye Data through Microsoft.

Related solution: semantic search over specialist documents

Specialist corpora break ordinary search. Legal, medical, scientific and technical material all have the same characteristic: the vocabulary is precise, it varies between authors and over time, and the person searching often knows the concept but not the phrasing. Keyword matching returns either nothing or everything.

ThirdEye Data builds the layer that understands the material rather than the words. That involves grouping documents by subject as they actually are, connecting the concepts the organisation uses to the language in the text, and giving each result enough summarised context for someone to judge it from a list. We build the correction path as well, because in specialist domains the experts using the system know more than the model does, and a system that cannot learn from them stops improving on the day it launches. We have built this for European case law, for a corpus of several million research documents, and for government operational reporting.

Explore our semantic search and document intelligence solutions ›