Title: Shared features of L2 writing: Intergroup homogeneity and text classification. Abstract

Download 179.28 Kb.

1 2 3

Table 6
Precision and Recall Finding (Training and Test Set): L1 and L2 corpus with four indices

Training set
Text set	Precision	Recall	F1
L1	0.479	0.792	0.597
L2	0.941	0.793	0.861

Test set
Text set	Precision	Recall	F1
L1	0.462	0.806	0.587
L2	0.949	0.793	0.864

Table 7
Accuracy of classifying text based on first language (four indices)
First language	English	Czech	Finnish	German	Spanish
Training set	0.799	0.756	0.735	0.913	0.724
Test set	0.821	0.650	0.780	0.961	0.672

Table 8
Differences in linguistic features between an L1 and L2 writer
Paragraph type	Writer background	Hypernymy	Polysemy	Stem overlap	Lexical diversity (m)
Introduction	English	1.790	4.258	1.000	0.022
	Spanish	1.490	2.976	0.000	0.013
Conclusion	English	2.094	5.241	0.857	0.023
	Spanish	1.427	2.804	0.200	0.018

1 A lower hypernymy score relates to less specific words, while a higher score relates to more specific words.

2 Maas, as reported by Coh-Metrix, is reverse-scaled so that lower numbers indicated greater lexical diversity.

Download 179.28 Kb.

Share with your friends:

1 2 3