Analysis of communication of uncertainty in genetic counseling patient letters for design of a natural language generation system

Abstract

The GenIE (Genetics Information Expression) Assistant is a proof-of-concept computer system designed to assist healthcare providers by creating editable first drafts of genetic counseling patient letters. Due to the wide range of genetic conditions to be covered and the rapid growth of knowledge in genomic medicine, it would not be feasible to have human experts to pre-author all possible text that might be needed. Thus, our approach has been to develop techniques for natural language generation of patient letters. To gain insight for the design of the system, a corpus of genetic counseling patient letters was studied. This article describes the methods used to analyze expressions of uncertainty in the corpus, some results of that analysis, and its implications for natural language generation.

Argumentation schemes: From genetics to international relations to environmental science policy to AI ethics

“…Like the Plan Deception scheme, these schemes are domain-specific but are related to AI plan recognition In addition, we identified the following schemes for arguing in faof a action of the protagonist: Reasoning, Resist Coercion, and Avoid Negative Consequences. 16 These latter schemes are related to more abstract schemes such as practical reasoning, argument from threat, and argument from negative consequences, respectively. (In subsequent work [25] in this domain, we proposed an Intentional Cause to Effect argument scheme and its critical questions for reasoning about future events resulting from an agent's intentional acts.…”

“…Although small in comparison to corpora used for machine learning, the corpus was representative of the genre and contained a variety of argument patterns 4. Although not incorporated into the final GenIE system, we also analyzed the different uses of probability statements in arguments in the corpus[15,16].…”

“…The schemes were based on an in-depth analysis of a 33-paragraph article[62] and a reading of several others 16. A kind of summary can be provided by listing the names of the schemes that were used from beginning to end of the Weinberger[62] article: Plan Distraction, Coercion, critical question of Coercion, Resist Coercion, Plan Deception, Inferred Plan, Coercion, Increasing Boldness, Coercion, Inferred Plan, Inferred Plan, Resist Coercion, Practical Reasoning, Avoid Negative Consequences, Avoid Negative Consequences, Practical Reasoning.…”

Towards Creation of a Corpus for Argumentation Mining the Biomedical Genetics Research Literature

“…Formal argumentation schemes describing arguments found in the corpus were defined (Green et al, 2011). Analyses of pragmatic features included rhetorical relations (Green, 2010a), ordering constraints and discourse markers (Green et al, 2011), point of view (Green 2005b), and use of probability expressions (Green 2010b). However, it was not a goal of that project to provide a publicly available corpus.…”