shlomihod / deep-text-eval

Differnable Readability Measure Regularizer for Neural Network Automatic Text Simplification
24 stars 7 forks source link

READ–IT: Assessing Readability of Italian Texts with a View to Text Simplification #12

Closed shlomihod closed 6 years ago

shlomihod commented 6 years ago

https://aclanthology.info/pdf/W/W11/W11-2308.pdf

shlomihod commented 6 years ago

Both document and sentence scoring Newspapers corpus - one normal and the other simplified. Why newspapers? Don't want to have variations due to the genre.

Binary classification (easy vs. difficult)

Lexical and synthetic features

SVM Document classification accuracy: ~98% Sentence classification accuracy: ~78%

Idea: simple text -> simple sentence (vice versa is not working)

Feng (2009) - a set of cognitively motivated features

Remember: we should do corpus features analysis first of all. 1