Research On Evaluation Label Extraction Technology For Game Review

With the rise of social media,more users choose to express their views on various online platforms.Most of these expressions are displayed in the form of commentary data for others to browse and further explore.As one of the fast-growing types of web platforms in recent years,the game platform has also left a large amount of user game comment data,which contains important value.These data can be processed and analyzed by using relevant techniques of natural language processing.The research of evaluation label extraction technology for game reviews has important guiding significance for exploring user information and helping to improve user experience.The main task of this subject is : given a piece of game comment text information,the model automatically extracts the evaluation features rich in game features.The research content belongs to the field of fine-grained sentiment analysis.There are four sub-tasks: the extraction of evaluation objects and the evaluation of words.Extraction,collocation of evaluation objects and evaluation words,candidate label clustering.In previous studies,most of the solutions to such tasks use a linear,streamlined approach,from the first subtask to the last subtask step by step processing,layer by layer,and the processing process is lengthy and complicated..This paper proposes a neural network model based on multi-task learning,which can extract the evaluation objects and evaluation words at the same time.After obtaining the evaluation object and the evaluation word respectively,the candidate tag can be obtained by simple analysis.Finally,the K-means algorithm is used to cluster the candidate tags,and a representative evaluation tag is selected.The final presentation of this study is the WeChat applet,which displays the processed review data in a small program.The displayed information mainly includes the processed game comments and the extracted game evaluation tags.By analyzing the game reviews in the existing major game platforms,the design generates corresponding game tags to help users quickly understand the game type and game style.At the same time,the player comments are analyzed to reflect the satisfaction of the majority of players with a certain game and the player's game.The emotions involved help other users to understand a game more quickly and intuitively.Simply put,users can quickly and easily understand the basic information of the target game and the quality of the game through this small program.