透過您的圖書館登入
IP:3.139.62.103
  • 期刊
  • OpenAccess

人工智慧學習平台之教學應用反思-以法律華語文本為例

A Study on the Pedagogical Application of AI-Powered Chinese Platform and Reflection: A Case Study of Legal Chinese

摘要


近年人工智慧大量應用於生活、自然語言處理各層面,例如語音辨識、語法分析、自動摘要、文字探勘主題分析等應用層面上(周福強、曾金金,2005;邵軒磊、曾元顯,2018),然這些技術應用於華語教學現場之實際效用卻少有著墨。本文從華語教師角度出發,以專業華語(Chinese for Specific Purpose)教材中之法律華語文本為主要研究內容,透過商用「龎帝智慧中文學習平台」(簡稱PR)介面轉換,針對PR系統自動化於法律華語文本的分級結果(文本、詞彙),分析其是否符合華語教學期待及華語教材編寫原則,並據此提出相對應的簡易系統優化可行建議。初步結果發現該系統有中文斷詞、語法之誤判或標記困難,未能符合華語教材編寫原則的科學性、實用性及針對性。有鑑於此,本文以華語教師為本,就語言分析提出目前PR自動轉換文本教材存在的主要問題並提出建議:(1)斷詞功能應考量搭配詞選擇與專業詞彙標記問題,宜同時考量字、詞本位面向;(2)增強高共現搭配組合的準確度及語法定式的模組訓練;(3)針對近義詞組,可結合「語義框架概念」探討,分析各句式的主要高共現搭配詞(Gries & Stefanowitsch, 2004a),提升系統語法標記的正確性;(4)建置不同面向的專業華語詞庫及主題分級。

並列摘要


Artificial Intelligence (AI) is an emerging technology with cross-disciplinary applications in Natural Language Processing (NLP), e.g., speech recognition, grammar tagging, automatic abstracting and text mining (Chou & Tseng, 2005; Shao & Tseng, 2018). However, there is a lack of discussion in regards to the practical effectiveness of the application of AI in real teaching environments. This study adopts the perspectives of current TCSL teachers and utilizes a piece of criminal law text automatically tagged in Ponddy Reader (PR) in order to make preliminary observations regarding issues of word segmentation and grammar pattern detection. Based on the text converted by PR, we draw implications regarding whether the platform meets the expectations of TCSL and principles of CSL material compilation. This study aims at addressing these issues by proposing several solutions for optimization. Analyzing the criminal law text automatically converted by the "AI-Powered Chinese Learning Platform", we detected major problems as below: segmentation errors, gaps in grammar pattern identification, and violation of the scientific, practicability and targetability principles of CSL materials compilation. Concerning the main problems of automated abstracting existing nowadays, the recommendations of this study include: (1) collocating words should be considered when dealing with unknown words and word tagging to avoid over-tagging, aside from the combination of character-based and word-based approaches; (2) another way to improve the accuracy of grammar pattern detection is to strengthen the training in high collocational patterns and formulaic speech. (3) for enhancing the accuracy of near-synonym distinction, one plausible method is to incorporate semantic frames and collocational analysis to extract strongly collocated keywords (Gries & Stefanowitsch, 2004a). (4) developing a multi-dimensional database of Chinese for specific purposes vocabulary collected from different disciplines and themes with varying levels of difficulty.

參考文獻


何寶璋、羅雲(2012)。非目的語環境華語國別教材編撰的思考、設計與實施。華語文教學研究,9(3),31-46。
宋曜廷、陳茹玲、李宜憲、查日龢、曾厚強、林維駿、張道行、張國恩(2013)中文文本可讀性探討:指標選取、模型建立與效度驗證。中華心理學刊,55(1),75-106。
林千翔、張嘉惠、陳貞伶(2010)。結合長詞優先與序列標記之中文斷詞研究。中文計算語言學期刊,15(3-4),161-179。
邵軒磊、曾元顯(2018)。文字探勘技術輔助主題分析—以 [中國大陸研究] 期刊為例。問題與研究,57(1),29-62。
陳姿青、張靜怡、李宜年(2007)。教育科技於華語教學之運用。明道通識論叢,2,179- 202。

延伸閱讀