Research On Mobile Phone Users’ Demands Based On Online Danmu And Comments

With the rapid development of the Internet,new mediums are emerging and digitalization is flooding people’s daily lives.Users can get the information they need anytime and anywhere by watching videos,audio,live broadcasts,etc.At the same time,users are happy to give feedback on the information they get in the form of danmu and comments.This results in unstructured data such as danmu and comments.Danmu are short texts with user interactivity and real-time features that facilitate communication among users,while comments are medium-length texts that express users’ views in a general way.These two types of texts not only represent users’ opinions but also bring service providers more opportunities to explore users’ ideas and understand their needs better.However,short texts are complex to analyze because of the diversity of evaluation objects and the incomplete nature of their grammatical structure.And few studies combine danmu with comments for comprehensive analysis.Based on the above background,this paper uses danmu and comments of the cell phone evaluation videos in the bilibili.com,and uses topic model and fine-grained sentiment analysis,combined with the management theory of kano model,to build a model to achieve comprehensive information extraction of text data and to study user needs.Firstly,this paper classifies cell phones into four categories: cost-effective cell phones,mid-range cell phones,high-end cell phones,folding phone.And uses web crawler technology to crawl the danmu and comments contained in various cell phone evaluation videos in bilibili.com,crawling more than 290,000 danmu and 170,000 comments in total and pre-processing the obtained data.Second,since the LDA topic model is not suitable for short text and the danmu text has the characteristic of sparse structure,this paper proposes the UIE-bertopic framework for topic modeling.After extracting the information from danmu text,the superfluous danmu without entities are removed,and danmu topics are further extracted by using the bertopic topic clustering model.As for the comment text,this paper uses the most widely used LDA topic model for topic extraction.Then a topic fusion method is proposed to merge the topics of data from two perspectives: danmu and comments.Finally,this paper sets aspect words based on the theme extraction results,conducts fine-grained sentiment analysis,finds out the user attention and satisfaction,and explores the different demand focus of cost-effective cell phones,mid-range cell phones,high-end cell phones,and folding cell phone users for cell phone performance respectively combined with kano model theory.It is found that for cost-effective cell phone manufacturers should maintain the excellent performance of the three attributes of appearance,processor,and battery life,and selectively improve the screen,price,and system attributes according to the cost budget,and after-sales service belongs to the highlight of attracting users,which should be done well while forming a difference with other brands.For mid-range cell phone manufacturers should give priority to improving the processor,screen,range,cost performance,and appearance design and texture in their highlights.High-end cell phone manufacturers should maintain the system,processor,and product experience,to improve the screen,battery life,brand management,and product design to create a highlight.For folding screen cell phone manufacturers should maintain the appearance and product experience attributes,further enhance the screen performance,form differentiated brand management,and after-sales service and make good system highlights.