Automaatika referaat (eng)
human
raters themselves (Elliot, 2003; Landauer et al., 2003; Page & Petersen, 1995). All AES
engines have obtained exact agreements with humans as high as the mid-80's and
adjacent agreements in the mid-high 90's--slightly higher than the agreement coefficients
for trained human raters. Several validity studies have suggested that AES engines tap
the same construct as that being evaluated by human raters. Page, Keith, & LaVoie
(1995) examined the construct validity of AES, Keith (2003) summarized several
discriminant and true score validity studies of the technology, and Attali & Burstein
(2006) demonstrated the relationship between AES and instructional activities associated
with writing.
AES is not without its detractors. Ericcson & Haswell (2006) performed a
comprehensive critique of the technology from the perspective of those who teach
postsecondary writing
annaabi.ee 
