WriteAhead2: Mining Lexical Grammar Patterns for Assisted Writing

Abstract

This paper describes WriteAhead2, an interactive writing environment that provides lexical and grammatical suggestions for second language learners, and helps them write fluently and avoid common writing errors. The method involves learning phrase templates from dictionary examples, and extracting grammar patterns with example phrases from an academic corpus. At run-time, as the user types word after word, the actions trigger a list after list of suggestions. Each successive list contains grammar patterns and examples, most relevant to the half-baked sentence. WriteAhead2 facilitates steady, timely, and spot-on interactions between learner writers and relevant information for effective assisted writing. Preliminary experiments show that WriteAhead2 has the potential to induce better writing and improve writing skills.

Applying the Bundle–Move Connection Approach to the Development of an Online Writing Support Tool for Research Articles

“…One writing support tool, WriteAway (http://writeaway.nlpweb.org), for example, provides lexical and grammatical suggestions with phrase examples based on an academic corpus while a word is being typed. The rationale behind this design is that “learner writers sorely need concrete writing suggestions, right in the context of writing” and that “[l]earners could be more effectively assisted, if CALL tools provide such suggestions as learners write away” (Chang & Chang, , p. 106). Precisely the same reasoning applies to our new pedagogical tool, which is designed to facilitate the use and learning of move‐specific bundles in the research article genre.…”

A Design Space for Intelligent and Interactive Writing Assistants

“…While some models can be developed using very small data (between 1 to 100 examples) [10,85,158], the others require much larger data. If the training needs more data (around 100 to few thousands of examples) which is often the case for fine-tuned models, we categorize them as medium [37,97,255,256,271]. For larger datasets (around tens of 1000s of examples) we denote this as large [56,207,257].…”

TellMeWhy: Learning to Explain Corrective Feedback for Second Language Learners

“…In the second stage of the learning algorithm (Step (2) in Figure 2), we extract reference information. First, we extract grammar patterns from Corpus of Contemporary American English (COCA 1 ) using an existing method (in Figure 3) described in Chang and Chang (2015). Subsequently, we store examples corresponded to a keyword's grammar patterns (e.g., (discuss, V n): issue, topic, matters).…”