自動化知識概念結構之建構－以智慧穿戴裝置為例

Automatic Construction for Knowledge Structure - Using Wearable Device as an Example

李來錫(Lai-Hsi Lee) ； 陳韋翰(We-Han Chen)

《智慧科技與應用統計學報》 18卷2期 (2021/02) Pp. 20-40

摘要

過去知識概念分析在資料篩選上大多以人工進行樣本抽樣與知識結構判讀，但資料在人工辨識的過程中難免會產生誤差，再加上人工無法進行大量判讀。為了改善上述問題，本研究編寫程式以判讀並篩選資料的各項知識概念，並計算知識概念在各討論區下的權重，以了解消費者在討論時對特定概念的正確頻率。研究以穿戴裝置之討論為例，資料來源為Mobile01論壇和批踢踢討論區，透過網路爬蟲的方式抓取4824篇文章的內容與留言，並使用結巴（Jieba）工具對文字進行分詞，以增加判讀的精確度，並整理出6項知識概念，以TF-IDF各指標計算各項知識概念權重和關聯權重，最後確立適當指標並建構出知識結構並加以分析。

關鍵字

知識結構 ； 文件頻率 ； 知識管理

並列摘要

In the past, knowledge concept analysis mostly used manual sampling and knowledge structure interpretation for data selection, but the data will inevitably produce errors in the process of manual identification. And a large amount of manual data interpretation is hard to achieve. In order to improve the above problems, this study is to construct a program to automatically interpret and calculate the weight of knowledge concepts in each discussion area and to understand the correct frequency of consumers' specific concepts. Data source are the Mobile01 forum and the PTT. The program crawls the contents of 4824 articles through the Web Crawler, and uses Jieba tool to segment the text. In order to increase the accuracy of the interpretation, and sort out 6 knowledge concepts, TF-IDF is used to calculate the weight and correlation weight of each knowledge concept, and finally construct and analyze the knowledge structure.

並列關鍵字

knowledge structure ； document frequency ； knowledge management