30 NOVEMBER 2024
ATOMICA
LABS
Sentiment
and
Time Series Analysis
PROJECT PROPOSAL
PRESENTED to:
Nabil Bin Hannan
Lecturer ECE Department
GROUP MEMBERS
Adrita Barua Moumi ID : 2211616042 Sec : 21
Md Farhan Jamal ID : 2111700642 Sec : 21
Md Raquib Hassan ID : 2022483642 Sec : 21
Web Scraping: Collect data
Adrita Barua Moumi ID : 2211616042
from the social media
Md Farhan Jamal ID : 2111700642 Web App Development
Machine Learning model
Md Raquib Hassan ID : 2022483642
and fine tuning it
Natural Language Processing is evolving day by day. Future Ai depends on
how we process and extract data from our past and then use it for a better
future to serve and secure humankind for a better future. Human mind is
more complex than
any-other thing. NLP enables computers and digital devices to recognize,
understand, and generate text and speech by combining computational
linguistics—the rule-based modeling of human language—with statistical
modeling, machine learning (ML) and deep learning.
We are doing a project with the aim of helping businesses. Sometimes it’s
hard to get the consumer's reaction and their opinion about the product.
For the proper growth of the business owners need to keep in touch with
the consumer's sentiment on the product. And make future decisions
based on that. It is hard to make a ‘Hero Product’ for the businesses and
maintain the quality with consumer’s demand. For doing so, sentiment
analysis is the most crucial part.
In today's competitive market, businesses must continuously understand
customer opinions and feedback to improve their products, services, and
overall brand perception. However, analyzing vast amounts of textual
data, such as customer reviews, social media comments, and survey
responses, is challenging and time-consuming. This limits businesses'
ability to act promptly on customer sentiments, potentially leading to lost
opportunities and dissatisfied customers.
Our project addresses this challenge by developing a Sentiment Analysis
Web Application using Python Streamlit. This tool leverages natural
language processing (NLP) techniques to analyze customer feedback in
real time, categorize sentiments as positive, negative, or neutral, and
provide actionable insights. It also generate sentiments csv file that we
can use with our sales analysis.
With the project, we can analyze the consumer’s comments and
reactions on the product and extract valuable insights -
what consumers are suggesting
what is the overall sentiment
with a time-series analysis, what is the sentiment trend
what they are saying about the “Quality”
what they are saying about the “Price”
what they are consuming alternatively for the product
Goals for the final
quarter
95%
We are implementing more functionalities for
analyzing and comparing the data in our web
app.
ATOMICA LABS | PAGE 5
Problem faces during our
project :
Users often describe the same information with different terms, leading to
synonymy. Research indicates that people only agree on keywords for well-
known objects 20% of the time, resulting in missed relevant materials during
searches.
Polysemy Defined
Polysemy occurs when a single word has multiple meanings. Terms like 'saturn',
'jaguar', and 'chip' can refer to various concepts depending on context,
complicating sentiment analysis.
Challenges in Sentiment Analysis
The variability in word usage caused by synonymy and polysemy presents
significant hurdles in sentiment analysis and opinion mining, where
understanding user intent is crucial.
Stemming as a Solution
Stemming is a technique used to normalize variations of words by reducing
them to their morphological roots. While it can aid in information retrieval, it
does not resolve issues where related words lack morphological connections,
such as 'physician' and 'doctor'.
We also face Slow inference, challenging for real-time applications, Prone to
overfitting on small or imbalanced datasets but we solve these problems with
the help of others
How can Sentiment & Time series analysis
help the business’s to grow beyond
Main Objectives:
To understand how user comments, reviews, and ratings evolved over
time and identify any significant trends or patterns.
Monitor public sentiment and analyze them
To provide valuable, data-driven recommendations for decision-
makers based on the extracted insights from the data.
Utilizing time series forecasting techniques to predict future rating
trends, enabling proactive decision-making.
1 [Combine sales trend]
2 [Making Reports]
Future Work Plan
3 [Analyze &train with larger ML models]
We have to do the projects into the proper steps. Our
initial steps will be-
4 [Performance test]
By simplifying sentiment analysis, our solution empowers businesses
to make data-driven decisions, enhance customer satisfaction, and
maintain a competitive edge.
ATOMICA LABS | PAGE 6
Contact us
www.sentiment.com
raquib.hassan@northsouth.edu
01990934865
