Augmenting the kappa statistic to determine interannotator reliability for multiply labeled data points
Andrew Rosenberg, E. S. Binkowski
Abstract
Open-access reader
Andrew Rosenberg, E. S. Binkowski
Abstract
Open-access reader
This paper describes a method for evaluating interannotator reliability in an email corpus annotated for type (e.g., question, answer, social chat) when annotators are allowed to assign multiple labels to a message. An augmentation is proposed to Cohen's kappa statistic which permits all data to be included in the reliability measure and which further permits the identification of more or less reliably annotated data points.
OpenAlex reports 34 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
This paper describes a method for evaluating interannotator reliability in an email corpus annotated for type (e.g., question, answer, social chat) when annotators are allowed to assign multiple labels to a message. An augmentation is proposed to Cohen's kappa statistic which permits all data to be included in the reliability measure and which further permits the identification of more or less reliably annotated data points.
Key concepts: Cohen's kappa, Reliability (semiconductor), Statistic, Computer science, Kappa, Identification (biology), Measure (data warehouse), Data mining