Research On User Comment Of Sweeping Robot Based On Text Mining

With the rapid development of artificial intelligence and other new technologies,intelligent products show explosive growth,home appliance scene intelligence has become an inevitable trend.In recent years,technological upgrading has brought technological dividends to the sweeping robot market,and the basic performance and intelligence degree of sweeping robot products have been continuously improved and optimized.More than90% of the world’s floor sweeping robots are made in China,which is also the world’s largest market by sales.The rapid innovation of enterprises in the industry,to maintain competitive advantages at the same time,the head of household appliances enterprises are also competing layout,research and development.Due to the particularity of online comments,a sentence usually contains different feelings for a variety of product attributes.Therefore,it is of certain research significance to construct a user-centered product attribute dimension system,obtain users’ emotional information for different attributes,and conduct application analysis on them.This thesis which named Research on User comment of sweeping robot based on text Mining,focuses on the e-commerce platform data collection,keyword extraction,text sentiment classification and topic extraction model of the relevant theories,jingdong website in the first half of 2021 market sales Top50 sweeping robot product reviews for empirical research.By web crawler technology,the comment data is obtained,and the data is cleaned and preprocessed.Combined with industry standards and user review data,using TF-IDF,Word2 Vec and other keyword extraction technologies,the product attribute dimension system is constructed,and the keyword clauses containing evaluation of specific attributes are extracted by regularization matching method.Based on the manual annotation data,deep learning methods such as LSTM(Long short term memory),Text CNN and BERT-LSTM were used to classify the keywords and phrases emotionally.And this thesis found that BERT-LSTM with the best evaluation effect,and selected it to classify the keywords and clauses emotionally.Based on the experimental data,the analysis and application of user reviews are conducted from the perspectives of enterprises and potential users,combined with different analysis needs and product dimensions,and relevant cognition is obtained.Firstly,the performance experience of sweeping robot is the primary attribute most discussed by users,among which the dust removal ability,material appearance and running sound are mentioned frequently.In terms of intelligent experience,most attributes mentioned by users are route planning,battery and energy saving,and obstacle crossing ability.Secondly,in terms of users’ complaints,attributes such as operating noise,floor washing ability,accessories,dust removal coverage,obstacle crossing ability are still the directions that need to be optimized and iterated.Thirdly,based on the perspective of potential users,the product is analyzed.In terms of dust removal ability attributes,brands on the market at present can be recognized by users.In terms of the attribute of washing ability,there are differences in the feedback of users of different brands;Haier and Mijia,two brands,are featured in the middle and low price market,and their performance experience and intelligent experience are at a disadvantage in user feedback.Fourth,based on experimental data,build data kanban to help enterprises quickly locate the sources of negative comments on specific products,and design quick selection tools to help users compare product attributes in a specific price range.