Influenced by dominant languages such as Japanese and Mandarin and transitions in history, politics, and society in Taiwan in the past one hundred years, one would expect to see rapid lexical changes in Taiwanese during this period. How many new words have been produced? How many words have disappeared? Are there rules or trends for speech change? This project intends to explore the answers to the above questions.We collected three versions of Taiwanese New Testament Bibles: ”The New Testament Sin-iok” (Barclay translation, published in 1916), ”The New Testament of Ko Tan version” (The red-covered Bible, published in 1972) and ”The New Testament Translated in Modern Taiwanese” (published in 2008), which are written in Romanized scripts (Pe̍h-oē-jī, vernacular writing)-as our corpora. We input these texts into a computer, tagged the metadata, then transcribed them into the mixed Han-Romanized script paragraph by paragraph. Next, we developed a program to align the above two scripts word by word, counted the word frequency, comparing words that appeared in the three versions. We looked for common words, words that are no longer in use and words that have recently emerged among these three versions. Then, by matching word pairs (in Romanized script and in mixed Han- Romanized script) which contain the same Romanized script but different mixed Han-Romanized script from the different versions, including words in different accents or with different pronunciations, we were able to more precisely compute the lexical changes in Taiwanese for the past hundred years. Moreover, we analyzed the tendency and rationale for Taiwanese lexical change.The research results show that the three versions have more than 500,000 word tokens and 12,140 word types respectively. Among them, 1,900 (15.7%) word types only appear in the 1916 edition of the New Testament, and these are rarely used nowadays. Another 2,039 (16.8%) word types appear only in the 2008 edition of the New Testament. These ”new words” show apparent influence of Mandarin. In addition, we found some word tokens with similar meanings changing from vernacular pronunciation to literary pronunciation.