Special Issue Article
Cross Domain Opinion Mining in Synonymically Structured Database
|Related article at Pubmed, Scholar Google|
Opinion mining aims at classifying sentiment data into polarity categories positive (or) negative.Opinion mining is the field of analyze the people’s opinions, sentiments, attitudes and emotions from written language. It has been important for many applications such as opinion summarization, opinion integration and review spam identification. On average, human process six articles per hour against the machine’s throughput of 10 per second. However, the opinion information is often unstructured and/or semi-structured data in the internet. Online product reviews are often unstructured, subjective, and hard to digest within a short time period. The main objective of our proposed work is to determine the human opinion from text written in the web page automatically. Sentiment classification aims to automatically predict sentiment polarity of users publishing product based sentiment data. Applying sentiment classifier results in poor performance because each domain using different sentiment word. In order to train a binary classifier from one or more domains we propose a method to overcome the problem of existing cross domain sentiment classification methods. First we create a synonym database for both source and target domains and perform pos tagging. A product based sentiment classification using spectral clustering algorithm to align the domain specific words from different domains into unified clusters for opinion classification is developed. Sentiment sensitivity is achieved with the help of synonym database by measuring the distributed similarity between the words. To investigate the effectiveness of our method, we have compared it with several algorithms and develop a robust and generic cross-domain sentiment classifier.