Sentence Annotation for Aspect-oriented Sentiment Analysis: A Lexicon based Approach with Marathi Movie Reviews

Aspect oriented sentiment analysis is a research area in natural language processing concerned with analyzing opinions at aspect/feature level. In this paper we present the construction of sentiment annotation system designed for aspect oriented sentiment analysis of Marathi movie reviews. The objective of the work is to create annotated dataset for opinion summarization system. For opinion summarization systems, it is better to use aspect based approach and build summaries around specific aspects and provide in depth analysis of the review. Since manual annotation of large data is difficult and time consuming, we try to automate the task using lexicon based approach. With the help of aspect and sentiment lexicons, we demonstrate a rule based annotation scheme designed for movie reviews domain in Marathi language. The annotated dataset is evaluated for aspect identification and aspect-sentiment pairing. The system reported an average F1 of 0.76 and 0.71 respectively for both the tasks. The proposed model can easily be adopted for sentiment analysis in other domains in Marathi language.