Peg Duncan on Twitter points to an article on Law.com by an English academic, Adam Wyner, "Legal Ontologies Spin a Semantic Web." (By the way, if you're not following Peg on Twitter, you should be.) I was curious because of my interest in legal research and because of the the flirtation with the semantic web that Google Squared and Wolfram/Alpha seem to represent.
Obviously — to me, at least — if computers are going to be able to respond in a sophisticated, i.e. more helpful, way to our queries about law, there needs to be an agreed-upon set of terms and relationships among those terms capable of describing ("capturing"?) important aspects of the discipline. This is what I understand to be an "ontology" within information science. To pick a simple example from Wyner's article:
Hale, Williams and Partners is a law firm; Joan Williams is a partner at Hale, Williams, and Partners. Given what we know about law firms, we infer that Joan Williams works at Hale, Williams and Partners and that Joan Williams is a lawyer. To us, these inferences are obvious, but for a computer to make them, the knowledge must use explicit, formal inference rules
This seems fairly straightforward (though the mischievous might like to cavil about "works at"). But things get evidently trickier when it comes to analyzing case law.
Wyner sets out, as examples, what he calls "elementary questions that can be answered by any legal professional, but not by a computer." Some of these are, indeed, easy to imagine an ontology handling well: Where was the case decided? Was it a case of first instance or on appeal? What legislation or case law was cited? Such "meta" or process facts are already meat and drink for the big legal databases.
Other questions in his list, however, strike me as being potentially problematic: that is, all "legal professionals" might not agree on the answers under all circumstances:
- What was the basis of the appeal?
- What were the legal issues at stake?
- What were the facts?
- What factors were relevant in making the decision?
This isn't the place to go into the perplexities of precedent, but it is important to acknowledge that they exist. Indeed, it is fussing these perplexities that sometimes constitutes the practice of law. Which brings me to my point: among the Thirteen Ways of Looking at the Law, one way sees it as a language game, where, simply put, interpretation is what the players do. This inevitable need for interpretation is fundamentally at odds with the semantic web's need for a fixed, more or less easily applied set of rules of analysis: it's a case of ontology meets hermeneutics, or, if you like, Humpty Dumpty meets Alice.
Of course, we can sneak up on precision. We can and should develop an ontology that can be used to mark up cases and legislation so that the more basic queries can be better answered by machines. But while we do this, we'll have to watch that as we approach the argumentation boundary we aren't fed simplification in the interests of expediency and profit, results, for instance, in which employees at LexisNexis or BigLaw pluck a phrase from a judgment and tag it as the ratio.
Far more difficult is legislation. To bring an ontology to bear on that would, I suppose, ideally require a shift in the way statutes are drafted, so that at a minimum, for example, definitions are always able to be located as such. This is, in a way, what a civil code has accomplished. But beyond that simple sort of restructuring of the parts ("if this, then that"), I'm unclear about how legal semanticists, if that's an appropriate way to describe Professor Wyner, would propose to construct an ontology for statute law.
One thing that should make the task of characterizing cases by issue or subject matter easier is that cases are so heavily "linked" to each other via citations and referencing. If you "teach" the application those cases that are seminal in each legal area, a cite to one of those cases could be a "vote" making it more likely that a given case falls into a particular type. You could get to work data mining cases to establish patterns that a computer could recognize.
That would probably be much more valuable than a mere count of keywords such as "corporation" or even "contract". Many cases include such words despite presenting no significant corporations or contract law issues.
I won't go into the choice of law and characterization issues that give even judges a hard time.
Last but not least – cases that address important procedural issues would probably be really tough to recognize.
To help me get my head around the semantic web, I hope to be going to meetings of the new Toronto group appropriately named the Toronto Semantic Web Meetup Group. The next meeting is the evening of June 23rd, on "The Semantics of Taxonomies: how taxonomies bring meaning to information." Register here. There is still room if anyone would like to join us. I'm going to come back to this blog post after I've attended that meeting. ;-)