研究課題/領域番号 |
17K00276
|
研究種目 |
基盤研究(C)
|
配分区分 | 基金 |
応募区分 | 一般 |
研究分野 |
ヒューマンインタフェース・インタラクション
|
研究機関 | 大阪府立大学 |
研究代表者 |
デンゲル アンドレアス 大阪府立大学, 研究推進機構, 客員教授 (00773574)
|
研究分担者 |
石丸 翔也 大阪府立大学, 研究推進機構, 客員研究員 (10788730)
黄瀬 浩一 大阪府立大学, 工学(系)研究科(研究院), 教授 (80224939)
|
研究期間 (年度) |
2017-04-01 – 2020-03-31
|
研究課題ステータス |
交付 (2018年度)
|
配分額 *注記 |
4,680千円 (直接経費: 3,600千円、間接経費: 1,080千円)
2019年度: 1,430千円 (直接経費: 1,100千円、間接経費: 330千円)
2018年度: 1,560千円 (直接経費: 1,200千円、間接経費: 360千円)
2017年度: 1,690千円 (直接経費: 1,300千円、間接経費: 390千円)
|
キーワード | eye tracking / readability / legibility / fixation / electrodermal activity / 人工知能 / 認知科学 |
研究実績の概要 |
In this research, we aim to evaluate the readability of a document on the basis of eye movements. Since it is not realistic to ask participants to read documents with eye tracking devices every time, we develop a system which synthesizes artificial eye movements on the document and utilizes them for the readability measurement.
As mentioned in the progress report of the last year, we focused on a fixation analysis in FY 2018. We proposed a method to estimate fixation durations of each word of a document without asking people to read the document with an eye tracker. We investigated the importance of word length as a feature on the assumption that an increase in word length and complexity would rather increase the fixation duration. In addition, the scope of word vectors (pre-trained word2vec model) from the dataset vocabulary was also utilized to determine the effects in fixation durations of the words. On a dataset including 19 participants reading five documents, we achieved to predict fixation durations on each word with the coefficient of determination R2 score 0.47.
|
現在までの達成度 (区分) |
現在までの達成度 (区分)
2: おおむね順調に進展している
理由
The main part of the research plan for FY 2017 is to generate artificial fixation durations on unknown documents. As listed above, we proposed a fixation duration estimation method and reported preliminary results. In addition, we conducted a preliminary experiment aiming to assessing the readability score of a document which is scheduled in FY2019. One Master student finished writing his thesis on this project. By considering them, we may state that we reached our goals.
|
今後の研究の推進方策 |
We are going to involve saccades to cover regressions (re-reading on a document) and skimming. Then, we combine approaches for fixation/saccade estimation to synthesize realistic eye movements. We will develop demo applications, evaluate the performance, and improve the system if we still have more time to investigate.
|