Providing Automatic Assistance for Academic English Writing: A Study on Nominalization and Lexical Choices
Student thesis: Doctoral Thesis
Related Research Unit(s)
Automatic paraphrasing has received much interest in the computational linguistics community in recent years. One type of paraphrasing is lexical substitution (McCarthy and Navigli, 2009), which replaces a word or short phrase with another. Paraphrasing can also involve manipulation of the clausal structure of a sentence, with a range of options that has been described as the “Cline of Metaphoricity” (Halliday and Matthiessen, 2014). According to Halliday, metaphoricity happens inside a sentence when the actions (i.e. verbs and adjectives) are converted into nouns, a process that is known as “nomianlization”. Sentences with multiple verb phrases are considered complex sentences. These complex clauses can be joined either paratactically (i.e. coordinate clauses) or hypotactically (i.e. subordinate clauses). Turning verbs into nouns through nominalization can turn a complex clause into a simple one (Halliday and Webster, 2004). Analysis of academic English corpora has shown that derived nouns and gerunds are very common in professional and academic English texts (Bhatia, 2014). The current study is an endeavor to design an assisted academic English writing system that automates the nominalization and lexical substitution in the ESL/EFL writings. The contributions of the system include an algorithm for noun generation using a neural language model, a model design for an assisted writing CALL system, mappings for nominalization and academic lexical substitution. The evaluations of the system performance in classroom and students’ formal writing show that the system has significant impact in providing assistance to the students in their writing.