Table 5 The weights of linguistic features in score prediction.

Lexical diversity

Moving average type token ratio per essay

Lexical density

# of lexical words (token)/# of words

Lexical sophistication

# of sophisticated word types / total # of words per essay

Mean dependency distance

\({\rm{MDD}}=\frac{1}{{\rm{n}}}{\sum }_{i=1}^{n}|{{\rm{DD}}}_{i}|\)

Mean length of clause

# of words / # of clauses

Verb phrases per T-unit

# of verb phrases / # of T-units

Clauses per T-unit

# of clauses / # of T-units

Dependent clauses per T-unit

# of dependent clauses / # of T-units

Complex nominals per T-unit

# of complex nominals / # of T-units

Adverbial clauses rate

# of adverbial clauses / # of clauses

Coordinate phrases rate

# of coordinate phrases / # of clauses

Semantic similarity

Synonym overlap/paragraph (topic)

Synonym overlap / paragraph (keywords)

word2vec cosine similarity

(a). word2vec → (b). cosine similarity between sample and reference

Metadiscourse marker rate

# of metadiscourse marker (type) / Number of words

Grammatical error rate

# of errors per essay