METHODS OF AUTOMATIC ASSESSMENT OF ESSAY TEXTS: THEORY AND CONCEPTUAL APPROACHES

The article deals with the development of a system of automatic evaluation of essay texts using modern methods of machine learning and natural language processing, directed to automating and improving the processes of evaluation of students' written work. Special attention is given to the application of various classification algorithms and deep neural networks such as random forest, logistic regression and recurrent neural networks including LSTM (Long Short Term Memory). The article describes in detail the data preprocessing process involving lemmatisation and tokenisation of the Tajik language, which underlines the importance of adapting text processing methods to the peculiarities of this language. The use of the TF-IDF algorithm to represent texts in numerical form is also discussed, which is an important stage of data preparation for model training. TensorFlow and Keras are used as the training platform. Taking into account the difficulties of working with the Tajik language, the authors present experimental results showing high accuracy of the model with MAE 3.47, which confirms the effectiveness of the proposed approach. It is expected that the elaborated system will increase the objectivity, accuracy and speed of evaluation of students' written work in educational institutions.

automatic assessment, machine learning, natural language processing, deep neural networks, tokenization

1. Inomov B. B., Tropmann-Frick M. Classification of scientific texts by specialty using machine learning methods // Bulletin of NSU. Series: Information technologies. – 2022. T. 20, No. 2. P. 27–36. DOI 10.25205/1818-7900-2022-20-2-27-36.

2. Maksudov Kh. T., Inomov B. B. / Assessing the effectiveness of k-nearest neighbors and logistic regression methods in determining the specialty of scientific texts // Polytechnic Bulletin series: Intellect. Innovation. Investments. – 2019. – 4(48). – Dushanbe: TTU, 2019. pp. 34–38.

3. Maksudov Kh. T., Inomov B. B., Mullojanov N. M. / Comparative analysis of the “decision tree” and “random forest” methods – when determining the specialty of scientific texts // Bulletin of the Tajik National University series: natural sciences. – 2019. – No. 3. – Dushanbe: TNU, pp. 23–28.

4. Tensorflow – Wikipedia URL: https://en.wikipedia.org/wiki/TensorFlow, accessed 2023-04-03.

5. TF-IDF – Wikipedia URL: https://ru.wikipedia.org/wiki/TF-IDF, access date 2023-04-03.

2026-03-26