Automatic Scaffolding and Measurement of Concept Mapping for EFL Students to Write Summaries.

Introduction A recent trend in evaluating students' language proficiency has been the assessment of their integrated reading and writing skills rather than corresponding discrete skills (Kirmtzi, 2009). University students who study English as a Foreign Language (EFL) are required to be equipped with both reading and writing skills in order to assist themselves in grasping main ideas or selecting key elements from reading a large quantity of texts in highly demanding academic courses (Dreyer & Nel, 2003). In acquiring reading and writing skills, students need to identify keywords or phrases from each paragraph, construct concept maps to decode texts, and integrate main ideas or key elements into organized and logical summaries (Kirmizi, 2009). Reading and writing success depends on word identification, cognition, construction, and summarizing skills (Sung, Chang, & Huang, 2008). Writing summaries is a particularly difficult task for EFL students to learn as they have to determine what content in a passage is the most important and then transform it into succinct statements in their own words (Yang, 2014). In determining what information is important in texts, the main idea of a passage is often not present in the surface structure (the exact wordings) of the text (Friend, 2001; Kintsch, 1998) and the cognitive process which converts surface structure to an understanding of a text is internal and largely unobservable in onsite instruction (Alfassi, 2004; Fischer, 2003). As such, EFL novice summary writers may encounter difficulties as incorporating source text information into their own writing in terms of low reading comprehension skills (Esmaeili, 2002; Plakans, 2009) and the restriction of their vocabulary size (Baba, 2009). To grasp the main ideas necessary for writing summaries, concept mapping has been reported to be an effective strategy which graphically indicates the relationships between multiple concepts (Tseng, Chang, Lou, Tan & Chiu, 2012). Liu, Chen and Chang (2010) have proposed that concept mapping strategies provide students with a more systematic and organized way to clarify the important concepts necessary for enhancing reading comprehension, thereby improving summarizing skills. In particular, concept mapping facilitates students externalize their prior knowledge and combine it with new ones for reconstructing new language knowledge and learning experiences (Novak, 1990; Hwang, Hung, Chen, & Liu, 2014). In light of the rapid growth of computer technology, computerized concept maps are helping to better students' reading comprehension and writing proficiency (Wu, Hwang, Milrad, Ke, & Huang, 2012). A concept map is a fill in-the-blank strategy, where students extract important words and phrases while reading a passage and then fill in the essential words or phrases in a map for comprehension and summary writing. Keyphrases, which include two or more keywords, are formed to represent important concepts (Mangina & Kilbride, 2008). Zha (2002) proposes that keyphrases and summaries comprise word-to-sentence relationships in terms of information retrieval generated by deleting unnecessary information, extracting keyphrases in each paragraph, and integrating keyphrases and main ideas to complete a summary. In forming concept maps, nodes refer to the main ideas, and links represent the associations between the main ideas (Adesope & Nesbit, 2013; Novak, 1993). A concept map is also a useful assessment tool for measuring students' reading comprehension and summary writing. Traditional paper-based concept maps cannot help the teacher immediately evaluate a student's comprehension. As a result, students are unable to receive timely feedback from their teacher or peers. With the assistance of technology, computerized concept maps facilitate the modification of nodes and links and make the task easier for students to fill in the keyphrases, revise their previous maps, and automatically receive feedback for scoring (Liu, Chen, & Chang, 2010). …