| 曾寰,龙小建,刘华.基于LDA及语义相似度的商品评论情感分类研究[J].井冈山大学自然版,2019,40(4):46-51 |
| 基于LDA及语义相似度的商品评论情感分类研究 |
| SENTIMENT CLASSIFICATION RESEARCH IN PRODUCT REVIEWS BASED ON LDA AND SEMANTIC SIMILARITY |
| 投稿时间：2019-01-18 修订日期：2019-05-10 |
| DOI：10.3969/j.issn.1674-8085.2019.04.009 |
| 中文关键词: 情感分类 LDA 语义相似度 观点词 |
| 英文关键词: sentiment classification LDA semantic similarity opinion word |
| 基金项目:吉安市社会科学研究项目（18GH113） |
|
| 摘要点击次数: 1572 |
| 全文下载次数: 1862 |
| 中文摘要: |
| 提出一种结合LDA及语义相似度的商品评论情感分类方法。该方法首先使用LDA对商品语料库建模，获取文档-主题矩阵；人工选择k对褒义词、贬义词，基于HowNet语义相似度计算主题（评价对象+观点内容）与各个褒义词和贬义词的相似度，达到对观点词极性判断，计算文本观点词情感极性的加权和作为文本的情感极性。实验表明，与基于向量空间的SVM分类方法相比，该情感分类方法在分类指标上表现更好。 |
| 英文摘要: |
| A product reviews classification method based on LDA and semantic similarity is proposed, The method firstly uses LDA to model the reviews corpus and get the review-topic matrix. Artificially, it selects k positive and negative words and use "HowNet" semantic similarity to calculate the similarity of opinion content words with each positive and negative words to get the polarity of opinion word. The weighted polarity sum of words will be the polarity of the review. The experiment result shows, compared with SVM classification method based on vector space text representation, the proposed method works better in classification index. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
