智慧手機客語拼音輸入法之研發──以臺灣海陸腔為例

Research and Implementation of Hakka Pinyin Input Method for Mobile Cell - An Example of Taiwan HioLiukAccent

根據客委會105年度所進行的全國客家人口暨語言基礎資料調查資料，全國最新客家人口比例為19.3%，總數為453萬7千多人，成為台灣第二大族群，因研究者本身為新竹客家人，新竹客語以海陸腔為主，故以海陸腔為基礎。本文開發智慧手機平台上客語拼音輸入法之海陸腔系統，在客語海陸腔輸入法中，除了基礎功能使用者偏好輸入、客語拼音字首快速輸入、前後詞預測、並提出了解決客語特殊字在智慧手機上無法顯示的問題之方法，讓客語特殊字以其拼音方式顯示。為讓完全無客語基礎之使用者也可以打出客語文章，我們增加中轉客輸入法，中文轉客語輸入法設計給毫無客語語言基礎或是尚在學習客語的使用者，讓使用者使用注音符號無聲調方式來輸入，輸入中文後、系統會自動比對其中文的客語詞彙，再進一步將輸入的中文轉成客語詞彙顯示。在系統客語語料庫中，包含客語單字拼音字庫5366筆、客語詞彙拼音字庫19351筆、客語前後詞字庫8505筆與中文客語對照字庫1679筆；語音合成系統縮需之語音檔有單詞2949個、詞彙1566個，這些資料為客委會和教育部所提供。在使用者評估調查中，在客語海陸腔輸入法的總平均4分，中文轉客語輸入法總平均分數為4.46分，客語語音總平均分數為4.3分，在台灣大多數使用者使用注音做拼音輸入，所以使用者的輸入習慣大大的影響其評估分數，在語音調查方面，使用者認為此系統有助於學習客語海陸腔的發音，也有助於數位學習的發展。

According to a survey data about the Hakka dialect made by Hakka Affairs Council in 2016, Hakka people in Taiwan is a population ratio of 19.3%, with 4.537 million peoples, making it the second largest cultural group in Taiwan. Coincidentally, the researcher of this thesis is a Taiwanese from Hsinchu city with a Hakka origin. Also, the Hakka accent in this area in Hsinchu is called HioLiuk accent. Thus, this thesis is based on HioLiuk accent. The Hakka input method not only has the basic functions such as Preference, Quick-type, words- prediction, but also displays a pinyin selection on the smart phone monitor for the user to key-in the desired word and instantly provides the special Hakka characters. This input method is designed for people who do not understand Hakka language well, and allows the users to use phonetic notation (without tones) as input. After keying in, the system will check its meaning and output the Hakka characters to finally choose from. There are 5,366 records in Hakka Single Word Pinyin database (table), 19,351 records in Hakka Word Pinyin database (table), 8,505 records in Hakka Previous and Successive Word database (table) and 1,679 records in Chinese To Hakka database (table). Speech synthesis system requires many voice files: 2,949 word files and 1,566 phrases provided by Hakka Affairs Council and Ministry of Education. In the survey questionnaire filled-out by users, HioLiuk input method got 4.0 points on average. Chinese To Hakka input method got 4.5 points on average, Hakka Language pronunciation got 4.3 points on average. Most users in Taiwan use pinyin input method, which has the big influence on the points. In the questionnaire about voice input, users believe that this system is conducive to those people who are learning Hakka and effective for the development of e-learning.