COLLEGE OF ENGINEERING AND
TECHNOLOGY
Department of Information
Technology
MAY, 2023
MATTU, OROMIA
Outlines
Introduction
Statement of the problem
Research Question
Objective
Scope and Limitation
Significance of the Study
Literature
Methodology
Introducton
Afan Oromo is a major indigenous African language that is widely
spoken and used in most parts of Ethiopia, as well as some
neighboring countries.
It is spoken by 40% of Ethiopians and also used as a lingua franca
among other groups who interact with the Oromo people.
Stance detection is a technique used to determine whether the
author of a text supports, opposes, or has no opinion about a
specific topic.
Detecting stance is important in social media, particularly for
analyzing public opinions on political and social issues
Cont.…
Stance detection plays an important role in analyzing public
opinion on controversial issues such as abortion, climate change,
feminism, referendums and elections.
Stance classification is a field of natural language processing that
has many uses.
It can recognize whether a community supports or opposes a
particular viewpoint related to religious and political topics
automatically.
It is also useful in developing recommendation and market
forecasting systems.
Statement of the problem
Internet have led to an increase in social media usage, allowing
people to express their perspectives and engage in idea exchange.
Many problems faced in the country such as displacement of
people due to civil war, unemployment, and high cost of living.
 These problems arise because the government does not
understand people's interests or viewpoints towards policies and
public interests.
Most research on analyzing people's attitudes on social media
regarding a particular topic has been done in English and
European languages
This is because it is hard to apply the existing models and
approaches to other languages like Oromic.
These problems arise because the government does not
understand people's interests or viewpoints towards policies
and public interests.
Most research on analyzing people's attitudes on social media
regarding a particular topic has been done in English and
European languages
This is because it is hard to apply the existing models and
approaches to other languages like Oromic.
Cont.…
Based on the above statement of problem, the research
questions are:
 Which Machine learning algorithm works best for
Afan Oromo stance classification?
 What is the impact of normalization on Afan Oromo
stance classification?
 To what extent the model performs in classify Afan
Oromo stance?
Research Questions
Objectives
General Objective
The main objective of this study is to develop an Oromic
stance classification Model using Machine learning
techniques.
Specific Objectives
 To achieve the main goal of the above general objective
the study will focus on several specific objectives.
Cont.…
 To review relevant previous studies to find suitable methods and
techniques.
 Collect and prepare data,
 Build and test different algorithms for classification,
 To develop an appropriate machine learning model for the
proposed work.
 To compare and evaluate the generated classification model using
different metrics.
Scope and limitation
Scope
Uses machine learning to detect the political stance in
Oromic language text
The researchers has going to collect posts and comments
from a Facebook page of a political party, annotated them
into two classes (in favor or against)
Limitation
The research has some limitations because there are not many other
studies to compare it with when detecting political opinions in
Oromic language text.
Can not determining who wrote a text and why they
have a certain stance,
Can not verifying if the person expressing their opinion
is legitimate.
Cont.…
Significance of the study
Used to measure public opinion towards events or entities.
Helps to understand the level of support related to social,
religious, and political topics by analyzing people's attitudes
and opinions towards them.
Used to identify people's political affiliations based on their
online posts during elections.
Used to detect whether someone supports or opposes
something based on their online posts
Literature Review
Authors Method Focus Area Observed Gap Result
Thiri Kyawa*, Sint
Sint Aungb
Machine learning Classification of the Stance in
Online Debates Using the
Dependency Relations
Feature
The model is done
for another language,
it is not for afan
oromo
Accuracy of 89.61%, 87.24%, 93.17%,
90.67%,
91.52%, and 88.44% for Abortion,
Creation, Gay
Rights, God, Gun
Rights, Healthcare
Ibrahim N. Awol1
and Sosina M.
Gashaw2
Three machine
learning
algorithms: Logistic
regression, Passive
Aggressive and
Decision tree
Lexicon-Stance Based
Amharic Fake News
Detection
They go for Amharic
,
Hybrid feature improves the area under
curve
from 0.982 to 0.995 by reducing the false
positive
Surafel Tadesse
Guda
machine learning
algorithms SVM,
LR and RF using
each feature
extraction
techniques
political stance detection on
amharic text using machine
learning.
They go for Amharic
,
Achieved accuracy score of 0.82 using
TF-IDF feature
extraction and SVM
Cont.…
Victoria
Yantseva 1,*
and
Kostiantyn
Kucher
machine
learning
Stance Classification
of Social Media
Texts for Under-
Resourced
Scenarios in Social
Sciences
It is done for
what ever text
of language ,
the author do
not define the
language, for
stance
classification
model achieving F1 macro of
0.76 on the test data for
a two-class classification
problem (negative/non-
negative stance for
immigration discussions).
These
Research
MACHINE
LEARNING
TECHINUQ
UES
STANCE
CLASSIFICATION
OF AFAN OROMO
TEXT USING
MACHINE
LEARNING
TECHINUQUES
Expected
Methodology
Research Design
 The reasearcher use DSR process to do the research.
 DSR has the following step.
 Problem identification
 Define objectives for a solution
 Design and development
 Demonstration
 Evaluation
Data preparation
 Data is collect from social media using a tool called
Facepager,
 The data is processed to remove unnecessary characters, split
sentences into words, remove punctuation and convert words
to their root form .
Performance Evaluation
 The accuracy of a classification model is going to
evaluate using four main metrics, those metrics are the
following:
 Accuracy,
 precision,
 Recall and
 F1-score.

Afan oromo stance classification using machine learning.pptx

  • 1.
    COLLEGE OF ENGINEERINGAND TECHNOLOGY Department of Information Technology MAY, 2023 MATTU, OROMIA
  • 3.
    Outlines Introduction Statement of theproblem Research Question Objective Scope and Limitation Significance of the Study Literature Methodology
  • 4.
    Introducton Afan Oromo isa major indigenous African language that is widely spoken and used in most parts of Ethiopia, as well as some neighboring countries. It is spoken by 40% of Ethiopians and also used as a lingua franca among other groups who interact with the Oromo people. Stance detection is a technique used to determine whether the author of a text supports, opposes, or has no opinion about a specific topic. Detecting stance is important in social media, particularly for analyzing public opinions on political and social issues
  • 5.
    Cont.… Stance detection playsan important role in analyzing public opinion on controversial issues such as abortion, climate change, feminism, referendums and elections. Stance classification is a field of natural language processing that has many uses. It can recognize whether a community supports or opposes a particular viewpoint related to religious and political topics automatically. It is also useful in developing recommendation and market forecasting systems.
  • 6.
    Statement of theproblem Internet have led to an increase in social media usage, allowing people to express their perspectives and engage in idea exchange. Many problems faced in the country such as displacement of people due to civil war, unemployment, and high cost of living.  These problems arise because the government does not understand people's interests or viewpoints towards policies and public interests. Most research on analyzing people's attitudes on social media regarding a particular topic has been done in English and European languages This is because it is hard to apply the existing models and approaches to other languages like Oromic.
  • 7.
    These problems arisebecause the government does not understand people's interests or viewpoints towards policies and public interests. Most research on analyzing people's attitudes on social media regarding a particular topic has been done in English and European languages This is because it is hard to apply the existing models and approaches to other languages like Oromic. Cont.…
  • 8.
    Based on theabove statement of problem, the research questions are:  Which Machine learning algorithm works best for Afan Oromo stance classification?  What is the impact of normalization on Afan Oromo stance classification?  To what extent the model performs in classify Afan Oromo stance? Research Questions
  • 9.
    Objectives General Objective The mainobjective of this study is to develop an Oromic stance classification Model using Machine learning techniques. Specific Objectives  To achieve the main goal of the above general objective the study will focus on several specific objectives.
  • 10.
    Cont.…  To reviewrelevant previous studies to find suitable methods and techniques.  Collect and prepare data,  Build and test different algorithms for classification,  To develop an appropriate machine learning model for the proposed work.  To compare and evaluate the generated classification model using different metrics.
  • 11.
    Scope and limitation Scope Usesmachine learning to detect the political stance in Oromic language text The researchers has going to collect posts and comments from a Facebook page of a political party, annotated them into two classes (in favor or against)
  • 12.
    Limitation The research hassome limitations because there are not many other studies to compare it with when detecting political opinions in Oromic language text. Can not determining who wrote a text and why they have a certain stance, Can not verifying if the person expressing their opinion is legitimate. Cont.…
  • 13.
    Significance of thestudy Used to measure public opinion towards events or entities. Helps to understand the level of support related to social, religious, and political topics by analyzing people's attitudes and opinions towards them. Used to identify people's political affiliations based on their online posts during elections. Used to detect whether someone supports or opposes something based on their online posts
  • 14.
    Literature Review Authors MethodFocus Area Observed Gap Result Thiri Kyawa*, Sint Sint Aungb Machine learning Classification of the Stance in Online Debates Using the Dependency Relations Feature The model is done for another language, it is not for afan oromo Accuracy of 89.61%, 87.24%, 93.17%, 90.67%, 91.52%, and 88.44% for Abortion, Creation, Gay Rights, God, Gun Rights, Healthcare Ibrahim N. Awol1 and Sosina M. Gashaw2 Three machine learning algorithms: Logistic regression, Passive Aggressive and Decision tree Lexicon-Stance Based Amharic Fake News Detection They go for Amharic , Hybrid feature improves the area under curve from 0.982 to 0.995 by reducing the false positive Surafel Tadesse Guda machine learning algorithms SVM, LR and RF using each feature extraction techniques political stance detection on amharic text using machine learning. They go for Amharic , Achieved accuracy score of 0.82 using TF-IDF feature extraction and SVM
  • 15.
    Cont.… Victoria Yantseva 1,* and Kostiantyn Kucher machine learning Stance Classification ofSocial Media Texts for Under- Resourced Scenarios in Social Sciences It is done for what ever text of language , the author do not define the language, for stance classification model achieving F1 macro of 0.76 on the test data for a two-class classification problem (negative/non- negative stance for immigration discussions). These Research MACHINE LEARNING TECHINUQ UES STANCE CLASSIFICATION OF AFAN OROMO TEXT USING MACHINE LEARNING TECHINUQUES Expected
  • 16.
    Methodology Research Design  Thereasearcher use DSR process to do the research.  DSR has the following step.  Problem identification  Define objectives for a solution  Design and development  Demonstration  Evaluation
  • 17.
    Data preparation  Datais collect from social media using a tool called Facepager,  The data is processed to remove unnecessary characters, split sentences into words, remove punctuation and convert words to their root form .
  • 18.
    Performance Evaluation  Theaccuracy of a classification model is going to evaluate using four main metrics, those metrics are the following:  Accuracy,  precision,  Recall and  F1-score.