Actor Identification and Relevance Filtering in Movie Reviews

With a large amount of data it is not always useful to run analyses on the entire corpus. Sometimes, it is helpful to previously preprocess data by filtering relevant information in order to form a fitting basis for the examination of particular aspects such as sentiment analysis. As a result, the amount of data that needs to be explored is reduced and concentrated, and thus the performance is enhanced. For example, a correct recognition of the rating of acting performances in movie reviews assumes that only judgements on the movie’s actors are used as a basis. In this paper, we discuss different approaches for a rulebased selection of sentences from movie reviews. Our aim is the filtering of sentences in order to facilitate analyses about single actors. Thereby actor identification is used to preselect a set of sentences that mention a specific actor. This is done individually for every actor involved in the movie. Furthermore, filtering is used to identify sentences that not only mention an actor but also state facts about him. To evaluate the developed methods, a test corpus consisting of ten movies with 30 reviews each, taken from the online movie platform IMDb, was built. Based on this data and the presented feature selection rules, an average F1 score of 77.9% is achieved as best result.