Post on 07-Aug-2015
TLS0070 Introduction to Legal Technology
Lecture 3 Artificial intelligence and law: the 21st century University of Turku Law School 2015-01-27 Anna Ronkainen @ronkaine anna.ronkainen@onomatics.com
Overall claim: Law is ~20 years behind other fields in intelligent tech adoption - nearby point of reference: language
technology - things originally considered AI don’t seem
all that impressive anymore (only annoying when not functioning properly): - spelling and grammar checking - speech recognition and generation - machine translation - ...
Why? - lawyers are conservative (but that’s too easy
an explanation) - lack of practically relevant research? - lack of commercial incentives - jurisdictional etc fragmentation means the
incentives are even smaller (but it’s the same for languages)
- law is HARD (but then you should just start with the low-hanging fruits)
How technologies change (or not): an example
What I worked on through much of law school... AnswerWizard/IntelliSearch, an intelligent tool for providing answers from on-line help files to questions posed in natural language, introduced in Microsoft Office 95:
But the next version (Office 97) might be more recognizable...
But the next version (Office 97–) might be more recognizable...
The basic tech was originally developed at the Stanford Research Institute (SRI)...
... and 10 years later, the same project gave us
The basic tech was originally developed at the Stanford Research Institute (SRI)...
... and 10 years later, the same project gave us Siri:
Another example: Watson: the Jeopardy-winning computer by IBM https://www.youtube.com/watch?v=lI-M7O_bRNg A different application https://www.youtube.com/watch?v=7g59PJxbGhY
DeepQA (Watson) high-level architecture
http://researcher.watson.ibm.com/researcher/view_group_subpage.php?id=2159
Watson merging and ranking algorithm
http://brenocon.com/watson_special_issue/14%20a%20framework%20for%20merging%20and%20ranking%20answers.pdf
...and we get
Modern approaches to legal AI: some examples
Putting it all together: From raw materials to Getting Things Done™ - Semantic Finlex: legislation as linked open
data - self-organized law systematics - recommender engine for law - INDIGO: intelligent backoffice processing for
public administration ...and plenty others (a task-based overview coming up at lectures 5–7)
Semantic Finlex - project carried out at Aalto U by Frosterus,
Tuominen, Hyvönen, funded by Tekes - Finnish legislation and case law as linked
open data - uses an ontology for legal source metadata
(which can be used to link them) - http://www.ldf.fi/dataset/finlex
(Frosterus et al 2014)
(Frosterus et al 2014)
Pros and cons - these kinds of resources are mandatory as
building blocks for more advanced things - it is available for Free™ - semantic enhancement only covers metadata
(not legal concepts, yet anyway) - based on 2012 legislation, no updates - only discovers explicit references
Systematizing Estonian laws through self-organization - project carried out at Tallinn U of Tech by
Täks et al - legal acts modelled as term vectors (based
on occurrences of individual words in each document) which are used to generate a self-organizing map (SOM, Kohonen)
- provides a 2-dimensional map of hypothetical (and also actual) relationships between statutes
(Täks & Lohk 2010)
(Täks & Lohk 2010)
Recommender engine for legal sources - project carried out at the Leibniz Centre at U of
Amsterdam by Winkels et al - uses networks of references (legislation ->
legislation, case law -> legislation) to find all documents matching the current document within a given horizon
- uses network topology based metrics to find the best matches (but plenty of other metrics to choose from)
- currently only prototype; in production could also learn from behavioural data (just like your favourite online store!)
Intelligent case management platform: INDiGO - project carried out for the Dutch
Immigration and Naturalization Service (IND) by Ordina, Accenture, Be Informed
- replaced an earlier paper-based administrative procedure
- intelligent decision support based on decision trees and checklists
- rules modelled in the system using a proprietary language
Semantic models in INDiGO - core taxonomies - regulations - online front office (UI) - catalog (index)
http://vimeo.com/43187024
Questions?