Modeling user attitudes toward a target from social media

A computer program product for user attitude modeling and behavior prediction for a social media network, the computer program product comprising a computer readable storage medium having program code embodied therewith, the program code executable by a processor to:

• collect data relating to previously demonstrated user attitudes, including sentiments, opinions, and actions, attributed to a plurality of social media network users toward a topic;
• create a model from the data, comprising:
• factorize the actions for behavior inference;
• factorize auxiliary content from the social media network for opinion and sentiment inferences, wherein the auxiliary content includes a user's profile data, content from the user's inputs, and aspects of the social media network;
• apply sentiment regularization and opinion regularization for respective sentiments and opinions to constrain user preferences attributable to the plurality of social media network users on implicit topics to explicit sentiments and explicit opinions;
• determine an overall sentiment, separately for each of the social media network users, from multiple opinions held by each of the social media network users with respect to the topic, the overall sentiment predicted by incorporating an opinion-sentiment transition matrix having two sentiment polarities into the model to capture relationships between multiple opinions held by one of the social media network users and corresponding sentiments;
• apply the model over a number of time segments to perform temporal analysis of the overall sentiment to provide a temporal evolution of user attitudes through an interface to another user;
• apply the model to a new user (u) of the social media network with respect to the topic;
• and
• generate a prediction of an attitude by the new user with respect to the topic comprising predicting sentiment and opinion of the new user as a function of the auxiliary content and feature coefficients learned during a training process, and predicting a future action of the new user as a function of the auxiliary content and latent profiles of the topic, wherein generating the prediction of the attitude comprises:
• predicting an opinion Ou of the new user using the model:
• Ou=F(Xu,W)=XuWt, where W is the feature coefficient that was learned from the model, and Xu represents a matrix of user-related features;
• predicting a sentiment Pu of the new user using the model:
• Pu=F(Xu,W)S=XuWTS, where the opinion-sentiment transition matrix S is learned from the model;
• and
• determining a probability of the new user taking an action on the topic:
• Ri,j=XuWTVTj, where a topic profile V is learned from the model.

A system for user attitude modeling and behavior prediction for a social media network, the system comprising:

• a memory having computer readable computer instructions;
• and
• a processor for executing the computer readable instructions, the computer readable instructions including:
• collecting data relating to previously demonstrated user attitudes, including sentiments, opinions, and actions, attributed to a plurality of social media network users toward a topic;
• creating a model from the data, comprising:
• factorizing the actions for behavior inference;
• factorizing auxiliary content from the social media network for opinion and sentiment inferences, wherein the auxiliary content includes a user's profile data, content from the user's inputs, and aspects of the social media network;
• applying sentiment regularization and opinion regularization for respective sentiments and opinions to constrain user preferences attributable to the plurality of social media network users on implicit topics to explicit sentiments and explicit opinions;
• determining an overall sentiment, separately for each of the social media network users, from multiple opinions held by each of the social media network users with respect to the topic, the overall sentiment predicted by incorporating an opinion-sentiment transition matrix having two sentiment polarities into the model to capture relationships between multiple opinions held by one of the social media network users and corresponding sentiments;
• applying the model over a number of time segments to perform temporal analysis of the overall sentiment to provide a temporal evolution of user attitudes through an interface to another user;
• applying the model to a new user (u) of the social media network with respect to the topic;
• and
• generating a prediction of an attitude by the new user with respect to the topic comprising predicting sentiment and opinion of the new user as a function of the auxiliary content and feature coefficients learned during a training process, and predicting a future action of the new user as a function of the auxiliary content and latent profiles of the topic, wherein generating the prediction of the attitude comprises:
• predicting an opinion Ou of the new user using the model:
• Ou=F(Xu,W)=XuWt, where W is the feature coefficient that was learned from the model, and Xu represents a matrix of user-related features;
• predicting a sentiment Pu of the new user using the model:
• Pu=F(Xu,W)S=XuWTS, where the opinion-sentiment transition matrix S is learned from the model;
• and
• determining a probability of the new user taking an action on the topic:
• Ri,j=XuWTVTj, where a topic profile V is learned from the model.