平板電腦化中文書寫表現評估系統之發展

Abstract

寫字是我國學齡孩童的主要職能之一，故書寫困難（handwriting difficulties, HDs）可能顯著學齡孩童的在校參與及課業表現，甚至造成其心理社會功能發展之長期障礙。有鑑於此，學齡孩童書寫表現之評估相當重要。然而，現有的評估工具在客觀性、全面性和適用性方面仍顯不足。因此，本研究旨在為台灣學童建立一平板電腦化書寫評估系統。 本研究的第一階段為建立功能性書寫表現的量化指標。文獻回顧顯示，書寫表現包含近端抄寫 (near-copy) 的書寫成果 (handwriting product) 和書寫歷程 (handwriting process) 兩大面向。本研究利用影像處理技術評估學童的書寫成果，並細分為準確度 (Accuracy)、字體工整度 (intra-character legibility)、及句子工整度 (inter-character legibility) 等量化指標。書寫歷程的評估則透過市售平板電腦內建的壓力感測功能監控筆尖在空中 (In-air movements) 和書寫表面 (On-surface movements) 的運動。第一階段研究已成功建立可客觀且全面性地評估學童中文書寫表現的量化指標。 第二階段的研究則著重於此系統之心理計量特性驗證。信度分析顯示字體工整度具有良好再測信度 (ICCs > .85) 以及可接受的隨機測量誤差 (MDC% <= 30.95%)。效度分析顯示本系統的字體工整度量化指標與國小教師的主觀易讀性評分達中至高度相關 (N = 75, rs = [-.33 – -.72])，且平板電腦為媒材的書寫評估可適當反映紙本書寫易讀性表現 (N = 30, rs = [.72, .57, .53] for Size, Orientation, and Deformation)。此外，本系統與台灣主流的紙本評估 (BCBL) 與問卷評估 (CHEF) 之中度關聯性亦符合檢驗建構效度的假設。第二階段研究顯示本系統具備一定程度的心理測量特性，可作為臨床或教育環境中的輔助評估。 本研究的最後階段則以全台各地共415位低年級孩童為基礎，建立全國性常模數據。由於書寫表現的量化指標非常多，本研究藉由點二系列相關 (point-biserial correlation) 及多變量變異數分析 (multivariate analysis of variance) 初步篩選出與抄寫困難最為相關的指標，並根據篩選過後的15項指標建立百分等級常模及標準分數常模。此外，由於性別顯著影響抄寫困難的發生率 (χ2 = 5.44*, p < .05)，故常模數據亦會根據性別進行划分。本研究後續以逐步多元邏輯迴歸 (Logistic regression) 釐清抄寫困難的關鍵指標，並建立綜合性的總結分數。迴歸分析顯示字體結構問題 (Deformation)、書寫動作控制 (Ratio of In-air length to On-surface length) 以及下筆力道 (Mean tip force) 對於抄寫困難的風險有直接影響力，三者之總結分數可有效區辨疑似抄寫困難的孩童。進一步分析接受者操作特徵(Receiver Operating Characteristic, ROC) 曲線亦揭示關鍵書寫指標的總結分數之最佳切截點為0.146（ROC曲線下面積0.80；敏銳度為76.67%；特異度為77.18%）。 本系統可客觀且綜合性地評估台灣孩童的中文書寫表現，心理計量特性的結果亦顯示其補充臨床評估，作為療效指標的潛力。此外，全國性常模和總結分數，亦可助於有效篩檢抄寫困難的學齡孩童。

Since substantial portions of Taiwanese children's time are allocated to handwriting, handwriting difficulties (HDs) could lead to long-term impacts on their academic and psychosocial development. Early identification of HDs help timely intervention. However, existing evaluation tools are insufficient in objectivity, comprehensiveness, and applicability. Therefore, this study aimed to establish a tablet-based Chinese handwriting evaluation system for Taiwanese school-aged children. In the first phase of this study, the system was preliminarily designed to address the challenges of quantitative handwriting performance. Two aspects of functional handwriting, namely handwriting product and handwriting process, were targeted. The handwriting product assessed what the written outcome looks like; the handwriting process addressed how children produced their handwriting. The domains of handwriting product included problems of accuracy, coherency (inter-character legibility), and construction (intra-character legibility) during copy tasks. The handwriting process assessed the In-air and On-surface movements (kinematics) of the pen-tip and the downward pressure of the pen (kinetics). With the utilization of image processing techniques and pressure-sensitive devices, objective, accurate, and comprehensive evaluation of handwriting performance was permitted. The second study phase focused on the psychometric examination. The reliability analyses (N = 25) demonstrated test reproducibility (ICCs > .85) and acceptable random bias (MDC% <= 30.95%) for intra-character legibility. The validity analyses demonstrated convergence with schoolteacher perceived legibility (N = 75, rs = [-.33 – -.72] for intra-character legibility) and moderate ecological validities between tablet and paper-based trials (N = 30, rs = [.72, .57, .53] for Size, Orientation, and Deformation). The expected relationships between our system and the readily available tool were also revealed as hypotheses testing for construct validity (N = 45, rs = [-.31 – -.38] for the paper-based BCBL, rs = [.30 – .50] for the questionnaire-based CHEF).Our system exhibited a certain extent of psychometrics, suggesting its complementary use in clinical and educational settings. To extend the applicability of our system, nationwide normative data were established in the final phase of this study based on 415 lower graders drawn from different regions of Taiwan. A set of 15 handwriting measurements that was most relevant to HDs was identified using the point-biserial correlation and multivariate analysis of variance. The nationwide norms, including percentile ranks and standard scores, were accordingly established for these measurements. Additionally, the prevalence of HDs was higher among boys than girls (χ2 = 5.44*, p < .05). Gender breakdown of these nationwide norms was included. Furthermore, through logistic regression, a summary score considering the collective effect of Deformation, Ratio of In-air length to On-surface length, and Mean tip force was developed to indicate the risk of HDs presence. The cut-off value of .146 of the summary score was also determined, yielding a sensitivity of 76.67% and a specificity of 77.18% for screening children with HDs. In conclusion, this study successfully developed a tablet-based evaluation of Chinese handwriting. Our system generally demonstrated good test-retest reliabilities and fair-to-moderate validities, suggesting its potential to complement clinical acumen. Furthermore, with nationwide normative data and the summary score, children with HDs could be effectively screened.