Argumentation Mining from Judicial Decisions: The Attribution Problem and the Need for Legal Discourse Models, Vern R. Walker, Parisa Bagheri and Andrew J. Lauria. Peer-reviewed and presented at the 2015 Workshop on Automated Detection, Extraction and Analysis of Semantic Information in Legal Texts (ASAIL 2015), held in conjunction with the 15th International Conference on Artificial Intelligence & Law (ICAIL 2015), in San Diego, June 8-12, 2015. Copy of full text available for download in .pdf format.
This paper discusses an attribution problem that is particularly critical for argumentation mining in judicial decisions, and the development of a legal discourse model as an important component of any solution to that problem. Attribution in this context is the problem of determining who believes a stated proposition to be true. This is a particularly difficult problem for developing natural language processing software that can use linguistic cues to automatically formulate and test hypotheses about who treats or accepts an expressed proposition as (probably) true, or relies upon or uses it as support. The attribution problem for argumentation mining from judicial decisions arises because the author of the decision (the judge) does not always believe the propositional content expressed by every sentence she writes in the decision. For example, a judge might write the sentence the varicella vaccine can cause neuropathy in humans, but writing this sentence does not always indicate that the judge herself believes the stated proposition to be true. The sentence might report an allegation of a party in a legal pleading, or the testimony of an expert witness, or the text of a document exhibit, as well as (or in contrast to) a conclusion or finding of fact by the judge herself. This paper argues that solving the attribution problem for argumentation mining requires development of an adequate legal discourse model – that is, a data structure representing the actors in a legal proceeding and the argument-related information that is important for understanding the meaning of a judicial decision. This paper discusses some basic content for such a legal discourse model that would be useful in making attribution determinations, drawing upon vaccine-injury compensation decisions in the United States for examples and to formulate hypotheses. The paper also argues that adequate development of a legal discourse model requires empirical investigation of actual judicial decisions.