kinfe19/NLP-Sentiment-Analysis-for-Amharic-Language

# Sentiment Analysis for Amharic Language # Project Description Amharic is one of the 86 languages spoken in Ethiopia. It is the "working language" of the federal government. The goal of this project is to have a more accurate sentiment analysis for Amharic language. The data that I will be using is an annotated data (tweet_ID with label of Positive, Negative or Neutral). The potential value is that there isn't a robust sentiment classifier for low-resource languages such as Amharic, so by working on a model hopefully I will improve the accuracy of the model. ## Dataset The dataset I will be using is composed of 9.4k Amharic tweets that are pre-classified as Positive, Negative or neutral. Link to data & description ## MVP A more accurate classification/NLP model that predict new Amharic tweets as Positive, Negative or Neutral. Input: A tweet written in Amharic Output: Positive, Negative or Neutral ## Goals The Amharic language is one of the low-resource languages and there have been some researches done in sentiment analysis. The goal of this project is to have a more accurate sentiment analysis for Amharic sentences by improving upon techniques used by previous researchers and also using pre-trained models that have a similar language structure with Amharic but with a high resource by utilizing transfer learning. ## Stretch Goals - Add a feature for sarcasm detection ### Justification that the scope of your project is appropriate for 6 weeks with respect to learning curve, compute and storage resources In order to successfully implement this project, I will do more research on NLP models, measuring and improving accuracy of NLP models. I will be using Google Colab notebook for running my model with GPU runtime. # Articles and Research Papers ### Summary #1 This research focused on the development of an annotation tool for Amharic tweets and the sentiment analysis for the data gathered. Since the annotation tools that already exist don't support …