| 主权项 |
1. A system for aggregating opinions generated by one or more sources in relation to an object, comprising:
a processor; and a memory coupled with the processor, the memory comprising:
a first opinion comprising a plurality of tokens;a probabilistic model representing the first opinion; andinstructions that, when executed, cause the processor to:
identify observed object variables based on content of the first opinion, wherein the observed variables comprise at least one observed source variable associated with a source of the first opinion and at least one observed object variable associated with a subject of the first opinion;determine whether an aggregate opinion database comprises an opinion associated with the source of the first opinion;based on a determination that the first opinion is associated with an opinion in the aggregate opinion database associated with the source of the first opinion, update the first opinion with latent variables associated with, but not expressly included within, the opinion in the aggregate opinion database associated with the source of the first opinion;estimate latent variables associated with, but not expressly included within, the first opinion based on the observed variables using a maximum likelihood technique, where estimating the latent variables comprises implementing a Markov model to estimate a source presentation format that defines a sequence of the tokens of the first opinion and estimating at least one latent object variable associated with a subject of the first opinion;update the probabilistic model with the estimated latent variables such that the updated probabilistic model defines a probabilistic relationship between the first opinion, the observed variables, the source presentation format, and the at least one latent object variable;update the first opinion with the estimated latent variables based on the updated probabilistic model; andupdate an aggregate opinion database based on the updated first opinion. |
