Presenting our PoC on document search for the Norwegian Courts Administration at Booster 2019

David Skålid Amundsen
Compendium
Published in
2 min readMar 21, 2019

--

This was the first time I attended Booster, and the first time I attended an IT-conference. One of the nice things about Booster is that it has something for everyone — from developers to functional architects and project leads. I was only able to participate on the second and third day of the conference, but found the number of workshops very refreshing compared to the academic conferences I have been to previously.

On the first day of the conference, Elin Langholm talked about “proof of concepts” (PoCs) and how PoCs can be used to quickly check if an idea will create business value and, if you fail, you fail fast and safe without large costs. On the second day Elin and I gave a talk on a specific PoC we made for the Norwegian Courts Administration the summer of 2018 on intelligent search, and the lessons we learned on the way. You can watch a recording of our talk below.

Building an intelligent search PoC

Elin Langholm & David Skålid Amundsen (30 minutes short talk)

All verdicts from the Norwegian Supreme Court (Høyesterett) must be made public and searchable on lovdata.no, but there is no such requirement on verdicts from the District Court (Tingretten) and the Court of Appeals (Lagmannsretten). Only a small fraction of the verdicts from these two courts are published on Lovdata. In the early summer of 2018 a judge in the District Court approached us and asked if we could help them search among all verdicts, not only those that are published on Lovdata. And not much time passed from the idea was first discussed to we started working on a PoC. We wanted to show that we could not only make a traditional search solution, but also use machine learning to improve the ranking of the documents and personalise the ranking to the person performing the search — increasing the business value even further.

In our talk we presented how we worked internally, putting together people with different backgrounds and skill sets, and with the customer. We also discussed our solution and the technology behind it, showed a live demo, and talked about things we learned that can be useful for others making PoCs. Here is the recording of our talk.

--

--

David Skålid Amundsen
Compendium

I am a data science and search consultant. I enjoy solving challenging problems within machine learning and search, sometimes both at the same time.