SENTIMENT ANALYSIS WITH WORD EMBEDDING-THE CASE OF DOUBLE-TRACK EDUCATION SYSTEM IN GHANA
Oscar Blessed Deho, Millicent Akotam Agangiba · 2019
This paper applies the concept of sentiment analysis for the determination of polarities (positivity, neutrality or negativity) of sentiments borne in the views expressed by Ghanaians regarding the newly introduced double track system in Second Cycle Schools in Ghana. These views are sourced from tweets (twitter posts). Accurate analysis of sentiments depends largely on the context of word usage. Most sentiment analysis approaches however ignore context when predicting sentiments; thereby leading to loss of context. In this paper, the loss of context is avoided with the use of the concept of Word embedding. Word embedding is a context-preserving technique which embeds the contextual information of data in the form of vectors before analysis of sentiment is done. An overall model accuracy of 76% was achieved using this technique. Our model’s accuracy outdoes similar works such as Garg’s (2016) work with an accuracy of 72%. The results from this work may help the Ghana government to get well informed on how the citizenry reacted to the reform of the educational system as well as help those at the helm of affairs to know how to roll out policies in the near future.