Reporting bias and knowledge acquisition

TLDR: This paper questions the idea that the frequency with which people write about actions, outcomes, or properties is a reflection of real-world frequencies or the degree to which a property is characteristic of a class of individuals.

Abstract: Much work in knowledge extraction from text tacitly assumes that the frequency with which people write about actions, outcomes, or properties is a reflection of real-world frequencies or the degree to which a property is characteristic of a class of individuals. In this paper, we question this idea, examining the phenomenon of reporting bias and the challenge it poses for knowledge extraction. We conclude with discussion of approaches to learning commonsense knowledge from text despite this distortion.

Figures

Table 2: N-gram frequencies for various verbal events and the number of times Knext learns that A person may 〈x〉, including appropriate arguments, e.g., A person may hug a person.

Table 3:Miles Travelled, Crashes, andMiles/Crash are for travel in the United States in 2006 [31]. A plane crash is considered any event in which the plane was damaged. Teraword results are for the patterns car (crash |accident), motorcycle (crash |accident), and (airplane |plane) (crash |accident).

Table 1: N-gram frequencies for (his |her |my |your) 〈body part〉 and the number of times Knext learns A 〈body part〉may pertain to a person. Plurals are included when appropriate.