Article Preview
Top1. Introduction
Dietary analysis is a universal concept which involves natural science, social science and philosophy. It has an effect on many fields, such as nutrition, medicine etc. Traditional survey methods such as paper-based or online questionnaires, observation and face-to-face interviews have been widely employed (Cooke & Jane, 2005). Although these methods benefit from custom research designs and targeting selected respondents, the sheer effort of surveyors and cost of resources are consuming.
The rapid development of social media attracts a large number of users. A report of We Are Social 20151shows that number of Chinese users of social media has achieved 659 million. These users have been accustomed to express opinions in social media (Sharma & Lbansal, 2015). As one of the daily activities, social media contain massive users’ opinions or attitudes about dietary. Two examples of user generated content about dietary are shown in Table 1. It can be seen that both the two users express attitudes about their comment targets. Therefore, these user generated content can be used to analyze users’ dietary preferences, which may make up for the deficiency of traditional methods about sample size, and reduce periods of data acquisition and analysis.
Table 1. Examples of online reviews about dietary from Weibo.com
| No. | Review Content |
| 1 | 吃德州扒鸡,味道好的!(Eat Dezhou Braised Chicken, taste good!) |
| 2 | 酸甜可口的菠萝咕咾肉。超级美味佳肴 (Sweet and tasty pineapple Sweet and Sour Pork. A super appetizing delicacy.) |
In this paper, the authors detect dietary preferences with microblogs in Sina Weibo2, which is one of the most popular social media in China (Zhang et al., 2014). The authors first extract aspects about dietary and then identify sentiment polarities (positive or negative) of aspects and dishes. In order to obtain more fine-grained preferences, this paper compares the effects of gender and region on dietary preferences which were found to have significant impacts on user preferences (Ahn et al., 2011; Moore & Zhang, 2010). Finally, the authors analyze evolutions of users’ dietary preferences over time. Empirical results on 3, 975, 800 microblogs indicate that users are discontent with the status of dietary in China. They have higher levels of demands. Atmosphere and appearance are important aspects for users. Meanwhile, differences about dietary preferences are obvious between genders and users from different regions. Moreover, users’ dietary preferences are constantly changing over time with different processes, especially users from different regions. In addition, according to the aspect extraction results, the authors find word2vec method can improve extraction performance effectively, which indicates that contextual information is important in identifying dietary aspects.
The remainder of this paper is organized as follows. Section 2 review related work and section 3 present framework and key technologies. Experiments and experimental results are described in section 4. Then the authors discuss about our methods and Conclusions in the last section.
TopIn this paper, the authors use social media to mining users’ dietary preferences. There are two aspects of related works, namely users’ dietary preference mining and sentiment analysis.