Research And Implementation Of Text Visualization Algorithm Based On Semantic Mining And Force-oriented Layout

In this information era,social media platform,online shopping malls,and news release platforms,which are brought by the Internet,have spawned several distinct types of information.One of this is called social media text.This type of text has a scattered structure,a large volume and rich information.However,large proportion of this kind of information is redundant.Effective visualization of this type of text can help users to quickly perceive current hotspots and popular opinions,bringing significant meaning to society.Text visualization aims at presenting the semantic features contained in a large amount of texts to users in a effective way,including the frequency of words,the importance of words,the logical structure of the text,the clustering of topics in multiple texts,and the trend of dynamic changes in topics.Two typical visualization techniques are word cloud and word tree.They either focus on displaying word frequency,or focus on the change of topic flow.However,none of them can reflect the semantics of the text itself while perceiving the overall content.Therefore,this thesis proposes a new type of visual structure that can not only perceive the emotion trend,but also retain the semantic content of the unstructured text,and displays it to users in an intuitive and visual way.So that the information can be better perceived by public opinion analysts and ordinary users.The main work of this thesis includes two aspects.First,this thesis proposes a new model for thematic and semantic visualization of social media text.The model first performs word segmentation,semantic segmentation,and part-of-speech tagging on the text set,effectively extracts the keyword pool.Then it proposes a lightweight sentiment classification algorithm which is right suitable for this model to do sentiment classification on the data set.After that it generates the best set of sequential patterns between vocabularies,showing the semantic structure between vocabularies,and form a semantic tree.Finally,based on the force-oriented layout technique,the model proposes a layout algorithm suitable for this scenario,putting each sequence pattern in the best position,also pruning them to avoid overlap and clutter.The model shows the semantics both inside and outside the sequence patterns very well.This model can both maximize the semantic mining of text content and reflect the concentration frequency of mining content in text.Secondly,based on the text visualization model we mentioned,this thesis designs and implements a text visualization system.The system performs sentiment analysis on the text data,shows the popular sentiment trend of the text,and displays the main views under positive and negative emotions.The system contains a user managing module,data preprocessing module and a visual module.Based on the aesthetic principles and guidelines of linked data,the model presents the social media text very visually.