基於大數據語料自動生成之中文詞彙聯想與實驗常模之比較

Abstract

本研究旨在比較基於大數據語料所自動生成之中文詞彙聯想（或稱詞彙共現）與基於真人實驗所建構之聯想常模。我們將Pecina（2010）中的57種詞彙共現強度計算法應用於巨量文本中，產生八萬五千多個常見中文詞彙兩兩間的共現強度（或稱聯想強度）。This study aims to compare two types of word association – the lexical collocations automatically generated using very large corpora and the association norms established in human experiments. Using very large text corpora, we computed the lexical association (or also called collocation) strengths between 85,346 Chinese words using the 57 word association measures described in Pecina (2010). Henceforth, we call the word association thus generated as the collocation dictionary. In order to validate the psychological reality of the automatically-generated word association, the Chinese word association norms established by Chen (1999) was used as a benchmark. The Chen word association norms consist of 1,200 stimulus words. In the free association experiment, each stimulus word was presented to 200 college students who were asked to write down the first word they came up with. For each stimulus word, the number of associate tokens is thus 200, but the average number of associate types is 86.