Automatic Thai-Language Essay Scoring Using Neural Network and Latent Semantic Analysis

Abstract

In this research, a backpropagation neural network and Latent Semantic Analysis were used to assess the quality of Thai-language essays written by high school students in the subject matter of historical royal Thai literatures. Forty essays written in response to a question were each evaluated by high school teachers and assigned a human score. In the first experiment, we used raw term frequency vectors of the essays and their corresponding human scores to train the neural network and obtain the machine scores. In the second experiment, we pre-processed the raw term frequency vectors using Latent Semantic Analysis technique prior to feeding them to the neural network. The experimental results show that the addition of Latent Semantic Analysis technique improves scoring performance.

A Comprehensive Review of Automated Essay Scoring (AES) Research and Development

“…It is commonly regarded that the semantics indicators are used to justify the whole or partial essays' semantics similarity (Islam & Hoque, 2013;Omar & Mezher, 2016;Sendra et al, 2016;Landauer et al, 2000;Ghosh & Fatima, 2008). The common syntactic indicators used in AES are spelling checking, stemming, lemmatization, word segmentation (Loraksa & Peachavanish, 2007), n-gram (Islam & Hoque, 2013;Chen & Zhou, 2019;Xu et al, 2017), and normalization (Taghipour & Ng, 2016;Ratna et al, 2019a). These were the essays' surface features, and they are found to be useful in grading essays (Ong et al, 2011).…”

“…The workflow of machine learning framework reported in Taghipour and Ng (2016) is illustrated in Figure 4. Looking at the recent trends in AES, the machine learning framework is gaining popularity in recent years due to the efficacy of SVM (Ratna et al, 2019b;Ratna, et al, 2019a;Xu et al, 2017;Awaida et al, 2019;Chen & Li, 2018) and the ability to represent text context with word embedding (Liang et al, 2018;Taghipour & Ng, 2016) propelled by Artificial Neural Network (ANN) (Loraksa & Peachavanish, 2007;Taghipour & Ng, 2016;Dong & Zhang, 2016;Liang et al, 2018).…”

Automated Evaluation of Writing – 50 Years and Counting

“…This consideration does not appear in Page ( 1966), yet it is an active line of subsequent work. While most of the research we cited so far has been on English, various aspects of writing evaluation, e.g., annotation, detection of various types of errors, and building AWE systems, have been researched for a variety of languages: Song et al (2016), Rao et al (2017), Shiue et al (2017) worked with data in Chinese, Lorenzen et al (2019) in Danish, Berggren et al (2019) in Norwegian, Amorim and Veloso (2017) in Portuguese, Stymne et al (2017) in Swedish, Berkling (2018) and Weiss and Meurers (2019) in German, Mezher and Omar (2016) in Arabic, Kakkonen et al (2005) in Finnish, Loraksa and Peachavanish (2007) in Thai, Lemaire and Dessus (2001) in French, and Ishioka and Kameda (2006) in Japanese. The list is by no means exhaustive; see Flor and Cahill (2020) for a recent review.…”

A hybrid automatic scoring system for Arabic essays

“…Another third instructor was asked to mark an essay if the other two instructors show two marks different by more than one. Our training set is small in size compared to other data sets that are usually used with LSA [7,13,15] which is due to the lack of an Arabic scored datasets.…”

“…They had tested more than five datasets, and the best result they obtained was a correlation of 0.63 between the automatic scores and the manual ones, whereas the interhuman correlation was 0.59. Loraksa and Peachavanish [15] discussed considering the sequence of essay information in LSA representation by dividing an essay into a number of parts; each part was represented by a vector in the LSA space. These vectors were combined into one vector representing the whole essay.…”