Recently, our research team has been mentoring a master student from the University of Buea, Cameroon on her research topic entitled: “SENTIMENT ANALYSIS ON SOCIAL MEDIA WITH REAL-WORLD APPLICATION”. The thesis was to fulfill a Master of Engineering (M.Eng) Degree in Telecommunications and Network Engineering from the Faculty of Engineering and Technology.
The candidate Amah Ann Nyoh Nditah defended successfully on the 14th, November 2018 and the jury gave her A grade. CONGRATULATIONS!!
Hereafter is a summary of her research work:
The research contributes to the vibrant field of Sentiment Analysis and Natural Language
Processing (NLP). It explores the techniques and tools applied to Sentiment Analysis on
social media, and focuses the application of these on Facebook. Over the years, research
works on Sentiment Analysis has revolved around mostly Twitter Sentiment Analysis and
focused on English Language as main processing language. This work performs Sentiment
Analysis of the comments scrapped from public Cameroonian Facebook pages and takes into
consideration that these comments are most likely in English or French.
The question that this dissertation explores is; how can Sentiment Analysis be applied to the
Facebook pages of Cameroonian organizations (given the use of both English and French as
official languages)? Can the results got from Sentiment Analysis be useful in improving the
operations of an organization should they make use of Sentiment Analysis on their Facebook
pages? In other to explore these questions, this research makes use of NLP techniques and
Sentiment Analysis tools. Text pre-processing tasks including; spelling correction, slang
replacement, emoticon replacement, lemmatization etc. create a platform for the unstructured
text scrapped from Facebook to be classified. It further deploys a language specific
classification system to handle the issues of English and French comments. In addition to
other pre-processing techniques deployed by Sentiment Analysis, it builds a lexicon to handle
emojis included in comments.
This research makes use of the hybrid approach to Sentiment Analysis, it deploys a
combination of a lexicon based and learning-based classifiers to carry out Sentiment Analysis
on the unstructured data. The resulting text is tagged with detected polarity (‘positive’, ‘negative’ or ‘neutral’). The results are displayed in a manner useful to the Facebook Page owner.
DiGIT is also committed to a partnership with high schools and universities to contribute to cutting-edge research in the area of computer science and innovative technologies.
DiGIT Labs Team