Abstract:
Abstract
Today, methods for automatic opinion mining on online data are becoming increasingly relevant. Over the past few years, methods have been developed that can successfully and with a great degree of accuracy analyze the sentiment in opinions from digital English text. These developments enable research into prediction of sentiment. In many cases, opinions are hidden in long forum posts and blogs. It is very difficult for a human reader to find relevant sources, extract pertinent sentences, read them, summarize them and organize them into usable forms. An automated opinion mining and summarization system is thus needed.
In this paper, we present a combined approach of lexicon based and machine learning that automatically extracts opinions from Amharic text. Most research efforts in the area of opinion mining deal with English texts and little work with Amharic text. We use a combined approach that consists of two methods. First, lexicon based method is used to classify sentences into positive and negative category. Second, the resultant classified sentences used as training set for machine learning method which subsequently classifies some other sentences. Combining the methods is required for better classification performance without the involvement of human annotated data. Instead manual labeling, the training examples are given by the lexicon-based approach.
We demonstrate the feasibility of sentiment prediction on Amharic text by developing a model. The performance of the model is tested on manually labeled data. Tests are done using comments given on movies. Experimental results show that the proposed combined method performs well with an accuracy of 82.5%.
To investigate the effectiveness of the proposed method for polarity detection, we compared it to the state of the art baseline methods and the result is very promising. However, to perform better evaluation, we need to extend the domain knowledge base and corpus used for training purpose.
Keywords: Sentiment Classification, Combined Classification, Amharic Opinion Mining
