基于场感知分解机的五笔输入法-《计算机技术与发展》

文章信息/Info

Author(s):: LI Ze-nan; LIU Han-ming* ; HU Zhen-zhen; LI Zi; SIMA Shen; GUO Gang; School of Mathematics and Computer Science,Gannan Normal University,Ganzhou 341000,China

Keywords:: field-aware factorization machines; Wubi input method; recommendation system; character amnesia; easy-to-use

摘要:: 计算机技术在中国的普及,使得人们大量使用计算机输入文本,从而大大减少了汉字的书写。加上拼音等易用的汉字输入法占据主导地位,使人们对熟悉的字变得生疏,“ 提笔忘字” 非常普遍。五笔等字形编码的汉字输入法体现了汉字的书写,可有效减少“提笔忘字冶等现象,但易用性不高。研究把推荐系统中的场感知分解机与传统的五笔输入法相结合,解决了稀疏特征问题,并根据用户的历史数据,预测用户需求同时推送最可能的候选汉字,提高了第一候选字词推荐准确率,降低了使用难度。实验表明,该五笔输入法具有稳健的“ 推荐冶能力,第一候选字词推荐准确率达到 98. 91% ,显著优于现有输入法,并且准确率可随用户对字词使用次数的增加而提高,达到了改善用户体验、增加用户粘性的目的。

Abstract:: The popularization of computer technology in China has led to a large number of people using computers to input text,whichhas greatly reduced the writing of Chinese characters. In addition,the easy-to-use input methods such as Pinyin have dominated,makingpeople unfamiliar with many characters,which lead to a problem is Character amnesia. The Chinese character input methods with glyphencoding such as Wubi embody the writing of the characters,which can effectively alleviate the problem,but they are not easy to use.The field-aware factorization machine is combined with the Wubi input method to solve the sparse feature problem. Based on the historyuser data,the proposed method can push the best candidate Chinese characters to the user by predicting their demand to improve the recommendation accuracy of the first candidate and to reduce the difficulty using Wubi. The experiments show that the proposed method hasa robust recommendation with accuracy of 98. 91% for the first candidate,being greatly better than that of the existing methods,and theaccuracy increases while the selection of the words increase,improving user experience and prompting their stickiness.