[目的/意义] 信息技术的快速发展使得用户评论、患者症状等短文本数据量迅速增长，如何从短文本中挖掘有价值的信息成为文本分类的研究热点。[方法/过程] 以国内某医院各科室患者的病情症状数据为语料集，针对短症状文本包含语义信息不足的问题，从各科室症状词的重要度与关联度出发，将症状文本中低于设定症状词数量的文本作为语义增强对象，采用Word2Vec与基于概率的TF-IDF算法抽取各科室的若干典型症状关键词，将其补充到语义增强对象中形成新语料集，最后利用机器学习算法对症状文本进行分类。[结果/结论] 基于文章语义增强方法构造的新语料集，相较于原始语料集，在支持向量机（Support Vector Machine，SVM）、多项式朴素贝叶斯（Multinomial Naive Bayes，MNB）以及随机森林（Random Forest，RF）上的分类效果均有较大幅度提升，准确率分别提高约10%、9%、10%。
[Purpose/Significance] With the rapid development of information technology, the amount of short text data such as user comments and patient symptoms has grown rapidly. How to mine valuable information from short texts has become a research hotspot in text classification. [Method/Process] This paper took the case symptom dataset of patients in various departments of a hospital in China as the target for semantic information deficiency of short symptom text. Based on the importance and relevancy of symptom words in each department, the study took the symptom text with less than the number of set symptom words as the semantic enhancement object, extracted some keywords of typical symptoms from each department by using Word2Vec and probability-based TFIDF algorithm, and add them to the semantic enhancement object to form a new lexical data set. Finally, a machine learning algorithm was used to classify the text of symptoms. [Result/Conclusion] Compared with the original predictive set, the new corpus based on the semantic enhancement method proposed in this paper has a better classification effect on SVM, MNB and RF, with an accuracy of about 10%, 9% and 10%, respectively.
[1] 杨长辉. 基于特征扩展的短文本分类方法研究[D]. 天津:天津理工大学, 2022.
[2] REINSEL D, GANTZ J, RYDNING J. Data age 2025:the evolution of data to life-critical[R/OL].[2022-09-28]. https://www.idc.com.
[3] NIGAM K, MCCALLUM A K, THRUN S, et al. Text classification from labeled and unlabeled documents using EM[J]. Machine learning, 2000, 39(2/3):103-134.
[4] 徐彤阳, 尹凯. 基于深度学习的数字图书馆文本分类研究[J]. 情报科学, 2019, 37(10):13-19.
[5] 周国剑, 陈庆春, 类先富. 基于概率主题和文本互扩展的短文本分类技术[J]. 计算机与数字工程, 2020, 48(10):2430-2435.
[6] KIM K, CHUNG B, CHOI Y, et al. Language independent semantic kernels for short-text classification[J]. Expert systems with applications, 2014, 41(2):735-743.
[7] ALSMADI I, HOON G K. Term weighting scheme for shorttext classification:Twitter corpuses[J]. Neural computing and applications, 2019, 31(8):3819-3831.
[8] ZHANG S, LV X, TANG Y, et al. Movie short-text reviews sentiment analysis based on multiFeature fusion[C]//Proceedings of 2018 international conference on algorithms, computing and artificial intelligence. New York:Association for Computing Machinery, 2018:396-401.
[9] 闫盛枫. 融合词向量语义增强和DTM模型的公共政策文本时序建模与演化分析——以"大数据领域"为例[J]. 情报科学, 2021, 39(9):146-154.
[10] 范昊, 何灏. 融合上下文特征和BERT词嵌入的新闻标题分类研究[J]. 情报科学, 2022, 40(6):90-97.
[11] 谢星雨, 余本功. 基于MFFMB的电商评论文本分类研究[J]. 数据分析与知识发现, 2022, 6(1):101-112.
[12] 张玉洁, 白如江, 刘明月, 等. 融合语义联想和BERT的图情领域 SAO短文本分类研究[J].图书情报工作, 2021, 65(16):118-129.
[13] 宋宁远, 裴雷, 王春迎. 科学论文语义增强的研究进展与趋势研判[J]. 图书情报工作, 2021, 65(1):82-90.
[14] 黄雷. 基于多词TF-IDF算法的智能导医系统研究[D]. 郑州:郑州大学, 2015.
[15] 徐奕枫, 刘利军, 黄青松, 等. 智能导医系统中TF-IDF权重改进算法研究[J]. 计算机工程与应用, 2017, 53(4):238-243.
[16] 孙震, 冷伏海. 一种基于知识元变异的ESI研究前沿知识演进分析方法[J]. 图书情报工作, 2022, 66(2):136-148.
[17] 覃世安, 李法运. 文本分类中TF-IDF方法的改进研究[J]. 现代图书情报技术, 2013(10):27-30.
[18] 丁敬达, 陈一帆, 刘超, 等. 基于共词和Word2Vec加权向量的文献-主题语义匹配分析方法[J]. 图书情报工作, 2022, 66(12):108-116.
[19] 程雅倩, 黄玮, 金晓祥, 等. 5G环境下高校图书馆自媒体平台多标签文本分类方法研究[J]. 情报科学, 2022, 40(2):155-161.
