One task common to all notice-and-comment rulemaking is identifying substantive claims and arguments made in the comments by stakeholders and other members of the public. Extracting and summarizing this material may be helpful to internal decisionmaking; to produce the legally required public explanation of the final rule, it is essential. When comments are lengthy or numerous, natural language processing and machine learning techniques can help the rulewriter work more quickly and comprehensively. Even when a smaller volume of comment material is received, the ability to annotate relevant portions and store information about them in a way that permits retrieval and generation of reports can be useful to the agency, especially over time. We describe a prototype application for these purposes. The Workspace for Issue Categorization and Analysis (WICA) allows the rulewriter to create a list of relevant substantive categories and assign them to marked portions of comment text. She can then retrieve all instances of a given issue within the comment pool. Preliminary results of experiments that apply text categorization and active learning methods to comment sets suggest that these techniques can facilitate the marking and category assignment process in lengthy or numerous comment sets. WICA will incorporate these techniques. Other possible applications of WICA within the rulemaking process are discussed.
Bruce, Thomas R.; Cardie, Claire; Farina, Cynthia R.; and Purpura, Stephen, "Facilitating Issue Categorization & Analysis in Rulemaking" (2008). Cornell e-Rulemaking Initiative Publications. 5.