On Improving the Accuracy of Readability Classification using Insights from Second Language Acquisition

shlomihod / deep-text-eval

Differnable Readability Measure Regularizer for Neural Network Automatic Text Simplification

24 stars 7 forks source link

Closed shlomihod closed 6 years ago

shlomihod commented 6 years ago

maxseidler commented 6 years ago

Summary:

Text readability classification based on age groups:
- 7-8 years
- 8-9 years
- 9-10 years
- 11-14 years
- 14-16 years
they build the WeeBit corpus, couldn't find it online though
again, many hand-crafted features were extracted
the used an MLP (93.3% acc when used on all features)
- they don't describe any specifics / hyperparameters of their MLP though