講演抄録/キーワード |
講演名 |
2021-09-10 13:15
Predicting Code Reading Test Answers by Using Eye Movement Features ○YUE YAN・Minoru Nakayama(Tokyo Tech) ET2021-11 |
抄録 |
(和) |
(まだ登録されていません) |
(英) |
Code reading comprehension progress has been shown to distribute in the eye movement data, which makes predict the code reading test answer correctness through eye movements comes to possible.This paper proposed a model based on the Residual Network and Transformer model to predict code reading answers with cluster data preprocessing. The model as an end-to end model which it can extract features from the raw data and make predictions and the model is evaluated on Eye Movement in Programing dataset,207 participants of different programing levels read two different JAVA codes then choose one of the four answers and recorded eye movement data.
The experiment results shows that our approach improves the average accuracy from 65% to 76% with this data preprocessing method and 73% to 76 % with this model structure on four data sets. |
キーワード |
(和) |
/ / / / / / / |
(英) |
Code reading / Test answer prediction / Eye features / Deep neural network / / / / |
文献情報 |
信学技報, vol. 121, no. 174, ET2021-11, pp. 17-22, 2021年9月. |
資料番号 |
ET2021-11 |
発行日 |
2021-09-03 (ET) |
ISSN |
Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
PDFダウンロード |
ET2021-11 |