中国网络大众生育态度倾向变迁——兼论舆情大数据在人口学中的应用
The Transition of Fertility Intention Propensity of Chinese Web-Users: The Application of Public—— Opinion Big Data Analysis

摘要 文章探索了如何利用网络舆情大数据来推断大众生育态度倾向的变迁。基于对网络评论的简单情绪分类、LDA非监督分类和fasttext监督分类结果的比较，文章发现依托于计划行为理论框架的监督分类法能提供更准确也更有意义的分析结果。从趋势来看，自2012年以来，涉及生育的评论态度倾向从积极占多数快速转变为消极占主导，这个过程也伴随着话题维度的转变，从更多行为态度上的表达转变为主要陈述生育面临的客观限制。房价、子女教育、工作成为提及最多的限制条件。这些结果说明低生育意愿更多是受客观限制的影响，而非观念转变。分省市生育情绪与地方生育水平较强的相关性体现了生育态度倾向分析的有效性。文章借此进一步讨论了舆情大数据在人口学研究中的应用前景。
Abstract：This study explores how to use the online public opinion big data to infer the transition of fertility intention propensity of Chinese web-users. By comparing the results from the simple sentiment analysis, the LDA un-supervised classification, and the fasttext supervised classification, it is found that the supervised classification based on the theory of planned behavior yields more accurate and meaningful results. A change from the positive to negative dominant mood of fertility intention has occurred since 2012, during which there is a shift of topics. The comments regarding perceived controls have replaced those related to behavioral beliefs as the most popular ones for fertility. Housing price, children's education, and work are the most mentioned key words in the category of perceived controls. The strong correlation between fertility intention propensity and fertility level among Chinese provinces validates our fertility intention analyses. We also discuss the prospect of applying public opinion big data in demographic research