Thank you for sharing!

Your article was successfully shared with the contacts you provided.

At the workshop on the “Natural Language Engineering of Legal Argumentation,” Dec. 13, 2008, in Florence, Italy, research was presented that contributes to the development of an “intelligent” support tool in natural language for arguing about the law. Development of such a tool is at the forefront of research in artificial intelligence and law. The workshop was part of the 21st International Conference on Legal Knowledge and Information Systems (JURIX 2008). With an intelligent support tool for argumentation, legal professionals could propose legal arguments that are then automatically represented and evaluated by a computer. As with many research areas, it will take time for the empirical and theoretical work to be done, then turned into a commercial software system available to legal professionals. Nonetheless, these developments will begin to affect legal practice in the foreseeable future. In particular, legal knowledge and reasoning will become more transparent, efficient, systematic as well as accessible to the general public. Legal professionals and law firms that are aware of this emerging technology will have a competitive edge. In this article, we outline several of the high-level components of the tool, how they fit together and the likely effect on legal practice. While no automated support tool for legal argumentation exists, we outline the requirements, apply the analytic methods from logic, linguistics and computer science to attack the issues and use available software technologies to develop it. We approach legal argumentation in natural language as a software engineering task, breaking the large problem into sub-problems, each of which is analyzed until we can solve and implement it as a software component. The components are then recombined to solve more complex aspects of the whole problem. A prototype system is developed, followed by revision and update. Along the way, we develop component tools and improve our understanding of how to represent and reason with legal information. There are three key issues: the defeasibility of argumentation, natural language and the representation of the conceptualization of the law. Legal argument has several senses: a case argued in court; a statement of a rule that is countered by facts, by an alternative rule or by exceptions; or procedural issues such as assigning burden of proof or accepting evidence as admissible. In general, adversarial parties make a claim for their side of the argument and rebut the claim for the opposing side. The notion of rebuttal reveals that argumentation is defeasible, meaning that while one party claims some statement is true, some other party argues that it is not. This contrasts with mathematical logic, where a claim once proven to be true is always true. Over the last 15 years, research in artificial intelligence and law has developed formal systems to represent and automatically reason with defeasible legal arguments. This opens the way to software systems that can quickly and systematically draw inferences from complex arguments. The sort of computational power available to the physical sciences to analyze, model and simulate problems is coming to the law. Legal argumentation is expressed in natural language, which machines cannot yet use and reason with. To understand legal argumentation, we must first understand the language of the law. To do so, we gather a corpus of legal texts in electronic form (e.g. PDF or HTML) such as found in legal decisions and other sources. To analyze a large corpus (e.g., several millions of documents), we apply automated computational linguistic techniques of text mining and machine learning. This enables us to identify lexicons (word lists), syntactic and semantic patterns, discourse relations as well as inference relations and inconsistency (see RTE). A related issue is the construction of the corpus. Libraries and legal professionals have access to databases of legal information that can be searched. However, this is often in a form that is proprietary and does not allow unrestricted search capabilities (see LexisNexis or Westlaw). To make progress, we create open-source, free, electronic databases such as in the project public.resource.org. The more we know systematically about legal language and how inferences are drawn, the better able we are to formalize this into a tool. Being able to search for complex patterns in open-source databases will enable firms to gather and analyze legal information in a way currently unavailable; for example, automatically correlating case factors with decisions. A formal representation of the conceptualization of the law requires a range of ontologies. An ontology is an explicit, formal and general specification of a conceptualization of the properties of and relations between objects in a given domain. An ontology can represent: the judicial hierarchy, indicating the roles, properties and relationships among the courts; the typical structure of a legal firm; the main topic areas in the law, sub-areas and key works; and case law. It provides researchers with a common vocabulary and organization of information. Assumptions, common knowledge or presuppositions are made explicit. To develop an ontology, one consults experts and expert information sources. An ontology developed following Semantic Web guidelines represents an expert’s knowledge in a machine-readable form. Rich sets of rules (if-then structures) and automated reasoning systems can be applied to the ontology to derive inferences. To instantiate an ontology, one adds instances which have the properties and relations specified by the ontology, thus creating a database. For example, given an ontology of the U.S. federal court system, one could instantiate this with the instances of the courts. Such a database can be queried to produce a list of relevant items. With such an ontology and rule set, the knowledge-base of a firm can be encoded, tested and automated. If a lawyer leaves a firm, the lawyer’s knowledge need not be taken with them. As the knowledge-base changes, the ontology and rule set can be easily maintained. Complex reasoning tasks that are beyond the abilities of any person such as consistency tests can be carried out. Databases of legal texts, ontologies, rules and inference engines are crucial components for the tool introduced above. The inference engine is the reasoning component, so just a part of the tool. However, we require a natural language interface so that the legal professional can read and write arguments in natural language. The tool is something like an augmented word processor: as text is entered, computational linguistic techniques are applied to automatically parse the statement into its grammatical components, and then automatically translate the grammatical analysis into a formal logic of the argument. Automated inference is applied to this logical representation. Because natural language is rich and complex, we use a controlled language that has a restricted vocabulary (limited to several thousand words) and limited grammatical forms (for example, using active sentence forms but not passives). A predictive editor allows only grammatical sentences to be input, much like word-completion found in some word processors. Though restricted, such systems allow an expressive professional-level style of writing. One such open-source system is Attempto. Using an interface such as that of Attempo, each argument that is input is entered into a network of other arguments that have already been entered by other participants. As the arguments are entered, additional components of the tool automatically compare the new argument against previous arguments. As the tool is interactive, questions, suggestions and implications can be provided which would otherwise require human expert knowledge. Reasoning engines are applied to the formal network of arguments to infer the justified arguments. Given the natural language input capabilities, the tool has a low learning curve; a broad spectrum of lawyers and legal support staff could use it. Furthermore, the language of the law is “normalized.” This could support more consistent, straightforward understanding of legal arguments. While a natural language interface is key to entering arguments into a network of arguments and getting output that we understand, it is not helpful where the argument network becomes large and complex such as in a real legal dispute. A textual representation of the dispute is linear simply because language is sequential. In contrast, the dispute has a nonlinear structure since the disputants may argue later about claims made earlier. A graphic representation can help to visualise this nonlinear structure such as the following fragment of the representation of an ordinary legal case by professor Henry Prakken. Click image to enlarge Nonlinear structure of a legal case. Click image to enlarge. Each argument entered by a party is represented as a node. The relationships between the nodes — one argument rebuts another argument — are represented as directed arcs (lines with arrows). This creates a graphic network of interrelated arguments. Automated inference engines can reason with such graphs to infer winning arguments. As users switch between the graphic and verbal representation of the dispute, they pick out points to argue further about. One argumentation visualization system is Araucaria. Several advantages of a graphic representation are that: a particular portion of the argument and all the related pro and con points are clustered together and easily examined; the interconnections among points in an argument are made clear; many people find graphic reasoning easier to manipulate. We have outlined the components of a legal argumentation support tool — defeasibility, text-mining, ontologies, natural language processing and a graphic representation. We have indicated a range of competitive advantages to lawyers and law firms. In future articles we will discuss these and other topics in greater depth. Dr. Adam Zachary Wyner is affiliated with the department of computer science at University College London, London, United Kingdom. He has a Ph.D. in linguistics from Cornell University and a Ph.D. in computer science from King’s College London. He has published on topics in the syntax and semantics of natural language, as well as artificial intelligence and law concerning legal systems, language, logic and argumentation. For further information, see Dr. Wyner’s blog LanguageLogicLawSoftware. He can be contacted via e-mail at [email protected].

This content has been archived. It is available through our partners, LexisNexis® and Bloomberg Law.

To view this content, please continue to their sites.

Not a Lexis Advance® Subscriber?
Subscribe Now

Not a Bloomberg Law Subscriber?
Subscribe Now

Why am I seeing this?

LexisNexis® and Bloomberg Law are third party online distributors of the broad collection of current and archived versions of ALM's legal news publications. LexisNexis® and Bloomberg Law customers are able to access and use ALM's content, including content from the National Law Journal, The American Lawyer, Legaltech News, The New York Law Journal, and Corporate Counsel, as well as other sources of legal information.

For questions call 1-877-256-2472 or contact us at [email protected]


ALM Legal Publication Newsletters

Sign Up Today and Never Miss Another Story.

As part of your digital membership, you can sign up for an unlimited number of a wide range of complimentary newsletters. Visit your My Account page to make your selections. Get the timely legal news and critical analysis you cannot afford to miss. Tailored just for you. In your inbox. Every day.

Copyright © 2021 ALM Media Properties, LLC. All Rights Reserved.