DIGITAL LIBRARY
AN APPROACH FOR AUTOMATED EVALUATION OF ESSAY-WRITING IN SECOND LANGUAGE LEARNING
Tokyo College of Transport Studies (JAPAN)
About this paper:
Appears in: INTED2016 Proceedings
Publication year: 2016
Pages: 8193-8200
ISBN: 978-84-608-5617-7
ISSN: 2340-1079
doi: 10.21125/inted.2016.0919
Conference name: 10th International Technology, Education and Development Conference
Dates: 7-9 March, 2016
Location: Valencia, Spain
Abstract:
There have been developed many applications software since it is believed that e-learning systems are effective for L2 (second language) learning. Some applications are helpful for collaborative learning while some are for self-study. We have implemented a collaborative e-learning system for practicing essay-writing in English. Students join the online class which their teacher sets up and write an essay on the system. Each class is run synchronously and thus a member can read the essays other members in the same class write in real time. Our system has a virtual facilitator called BOT who behaves as a member. He or she could refer to the model BOT submits as well as tips extracted from the BNC (British National Corpus) and the COCA (Corpus of Contemporary American English) if their teacher set them. After correcting or modifying their own essays several times by peer reviews, they exit the class and go to next class. We hope that they could improve their writing skills as completing classes. How can they feel it? It has yet to be shown the automated evaluation method for essay-writing while methods that evaluate a sentence grammatically have been proposed in some studies. The goal of this paper is to show the integrated method that automates evaluation in order to effectively utilize the history of essay-writing and to encourage students to raise motivation. The proposed method provides evaluation from three perspectives; vocabulary size, grammatical complexity and story line. The students enlarge vocabulary size through reading essays and tips in the system. The evaluation is based on frequency information in corpus, like several studies. The experienced are apt to use not only extensive vocabulary but also complicated sentences. For the purpose of accurate evaluation it is necessary to consider grammatical structure. In our implementation the score of grammatical complexity is calculated by the depth of XML tree which Enju (English syntactic parser library) returns. We don’t mind error in grammar because it must be corrected through peer reviews. The most difficult evaluation in essay-writing class is story line. Assume that teachers give a theme in each class and the model to students. Using Word2Vec which gives a characteristic vector of a word in a corpus, we can get the distribution of sentence vectors. If a students’ story line is similar to the model, the distribution of it can be near that of the model. Thus we define the score of story line as the distance between two distances. Objective evaluation plays an important role in L2 learning. Our method can provide helpful evaluation for both teachers and students to utilize the history of essay-writing. In this paper we propose new evaluation method and also show the result of evaluation in actual classes.