Computerized Model of Assessment Procedures of Students’ Reading Capabilities Based on Abridgement Method Employing Two-Pointer Problem Solution and Dynamic Programming

Kazuo Akashi, Hibiki Ito, Kunihiko Takamatsu, Kenya Bannaka, Shotaro Imai, Sayaka Matsumoto, Katsuhiko Murakami, Yasuo Nakata, Tetsuhiro GozuInternational Journal of Learning Technologies and Learning Environments 2025 · DOI

Introduction

The assessment of reading comprehension is a cornerstone of educational evaluation, yet it remains one of the most resource-intensive tasks for educators. Traditional summarization requires manual grading that is often subjective and prone to "marker dependence," where different evaluators may assign varying scores to the same work. This paper presents a computerized model designed to automate the grading process using the "abridgement" method, leveraging two-pointer and dynamic programming algorithms to transform a 15-hour manual task into a matter of minutes.

The Abridgement Method

Originally proposed by Japanese linguist Susumu Ono, abridgement is a constrained form of summarization. In this method, students are required to create a summary that is strictly a subsequence of the original text. This means students can only delete characters; they cannot add, swap, or reorder them. While this constraint simplifies the "search space" for correct answers and makes grading more objective, manual verification of hundreds of submissions remains a significant logistical hurdle for teachers. This research focuses on automating that final verification and feedback loop.

Algorithmic Framework

To automate the assessment, the researchers implemented two primary computational strategies to compare student submissions against the original source text and a model answer.

1. Two-Pointer Solution

This linear approach ($O(|S| + |T|)$ complexity) uses two pointers to simultaneously traverse the original source ($S$) and the student's text ($T$). While computationally efficient and intuitive, it is relatively "brittle." If a student inadvertently inserts a single character—breaking the subsequence rule—the pointers may lose alignment, failing to recognize subsequent correct matches and resulting in an unfairly low score.

Algorithm for mapping student abridgements back to original text indices.

2. Dynamic Programming (DP) / Longest Common Subsequence (LCS)

To address the limitations of the two-pointer method, the researchers employed a DP-based approach ($O(|S| \cdot |T|)$ complexity) to identify the longest shared sequence between the two texts. This method is significantly more robust; even if a student includes "invalid" characters (insertions), the DP algorithm can skip the errors and identify the valid clusters of the abridgement, allowing for fairer partial credit.

Evaluation and Judging

The system utilizes a judging function to determine if a submission adheres to the strict subsequence rules of abridgement. This function calls the matching algorithms to verify the integrity of the student's work.

The judging function used to determine if a submission is a valid abridgement.

Once processed, the system outputs a Microsoft Word document for each student with a three-color highlighting scheme for immediate feedback:

  • Pink: Characters chosen by both the student and the model answer.
  • Yellow: Characters in the model answer that the student missed.
  • Turquoise: Characters the student chose that were not in the model answer.

Experimental Results

The model was tested using a real-world dataset of 370 student submissions from Kobe Tokiwa University, Japan. The source text was a Japanese passage concerning voting ages and juvenile law.

Efficiency Gains

The manual grading of these 370 submissions previously took a human coordinator approximately 15 hours. In contrast, the computerized two-pointer method processed the entire batch in just 63 seconds, while the DP approach took 190 seconds. Both represent a massive leap in productivity, allowing educators to focus on supporting students rather than administrative marking.

Accuracy and Fairness

The study found that 73% of students followed the rules perfectly. For the 27% who submitted "invalid" texts, the DP approach proved superior. On these invalid submissions, the DP method yielded a mean score of approximately 75%, compared to only 64% for the two-pointer method. The DP algorithm's ability to "merge" fragmented correct segments ensures that students are not overly penalized for minor technical errors.

Conclusion and Future Work

The computerized abridgement model provides an objective, high-speed alternative to manual reading assessment. While the DP approach is fairer for scoring, the authors suggest a "hybrid" future model that uses DP for assessment but applies a small "non-adherence penalty" to maintain the rigor of the abridgement rules. Additionally, future work will explore dedicated user interfaces to prevent invalid submissions at the source and adapt the model for space-delimited languages like English.

Summary updated

BibTeX
@article{akashi2025computerized,
  author = {Kazuo Akashi and Hibiki Ito and Kunihiko Takamatsu and Kenya Bannaka and Shotaro Imai and Sayaka Matsumoto and Katsuhiko Murakami and Yasuo Nakata and Tetsuhiro Gozu},
  title = {{Computerized Model of Assessment Procedures of Students’ Reading Capabilities Based on Abridgement Method Employing Two-Pointer Problem Solution and Dynamic Programming}},
  journal = {International Journal of Learning Technologies and Learning Environments},
  year = {2025},
  doi = {10.52731/ijltle.v8.i1.873},
}
Figure