Methods for automated controversy detection of content

A method for automatically detecting controversy in electronic content, comprising the steps of:

• gathering by a processor a collection of documents (DBG) from a content source;
• constructing by the processor a set of controversial documents (Dc) from the collection;
• formulating one or more language models from both the collection of documents (DBG) and the set of controversial documents (Dc), wherein a language model includes a controversial language model (LC) and a non-controversial language model (LNC);
• calculating by a processor the probability (P) of each document (D) being controversial P(D|C) and a probability of each document being non-controversial P(D|NC) using the controversial language model (LC) and the non-controversial language model (LNC);
• providing by the processor an equation to estimate P(C|D) representing a probability that D is controversial and P(NC|D) representing a probability that D is non-controversial, wherein the providing step includes the step of applying one or more mathematical rules to reduce an equation to the form:
• log P(D|C)−log P(D|NC)>α,
• wherein α represents a pre-computed threshold value.
• comparing the probabilities P(D|C) and P(D|NC) to α to determine whether the document (D) is controversial or non-controversial;
• and
• communicating by the processor to a user interface or an application program interface (API) whether or not the document is controversial.