KNS INSTITUE OF TECHNOLOGY

Affiliated to VTU, Belgaum & Approved by AICTE

Bangalore -560064

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING

Technical Seminar on

Sentiment Analysis Using LSTM

Presented By – Nameera Shiraz

USN – 1KN20CS025
Semester – VIII
 Abstract 01

Introduction 02

Table of Sentiment Analysis 03

Content LSTM 04

Conclusion 05

References 06
Abstract
Analyzing the big textual information manually is tougher and time-consuming. Sentiment
analysis is a automated process that uses computing (AI) to spot positive and negative
opinions from the text. Sentiment analysis is widely used for getting insights from social
media comments, survey responses, and merchandise reviews to create data-driven
decisions. Sentiment analysis systems are accustomed to add up to the unstructured text
by automating business processes and saving hours of manual processing
Introduction
It’s estimated that 80% of the world’s data is unstructured and not organized during a pre-
defined manner. Most of this comes from text data, like reviews, emails, chats, social
media, surveys and articles. These texts are usually difficult and time-consuming to
investigate and understand. The sentiment analysis system authorizes company to create
sense of this huge amount of unstructured text
Sentiment Analysis
Sentiment analysis is text classification to detect positive, and negative sentiments. Useful for
analyzing customer reviews, survey responses etc. Recurrent neural networks like LSTMs are
effective for text modelling due to their ability to capture context and long-range dependencies.

Understanding sentiment analysis architecture and training using LSTMs has many real-world
applications in NLP such as:

• Analyzing customer feedback and reviews for products/services

• Social media monitoring – understanding public sentiment on topics, events
• Analyzing survey responses to guide business decisions
• Chatbots – detecting user sentiment from conversations
• Political sentiment analysis from speeches, news
• Healthcare – gauging patient satisfaction from feedback
Challenges in Sentiment Analysis
Traditional sentiment analysis models struggle with contextual understanding and long-
range dependencies. They often fail to capture the nuances of language, leading to
inaccurate results. This is where LSTM comes into play.

Understanding LSTM
LSTM is designed to address the vanishing gradient problem in traditional RNNs. It can
retain long term dependencies and selectively forget or remember information. This
makes it well-suited for sentiment analysis tasks.
 Long short-term memory (LSTM)
neural network (RNN) architecture employed in the sphere of deep learning. Unlike standard
feedforward neural networks, LSTM has feedback connections. LSTM [5] networks are well-
suited to classifying, processing, and making predictions supported statistic data since there
may be lags of unknown duration between important events in a very statistic. LSTMs were
developed to accommodate the exploding and vanishing gradient problems that may be
encountered when training traditional RNNs.

Training LSTM for Sentiment Analysis

Training an LSTM model for sentiment analysis involves preprocessing text data, embedding
words into vectors, and finetuning the LSTM layers. The model learns to extract sentimentrelated
features from the input text.
LSTM Architecture
The layers are as follows:

0. Tokenize : This is not a layer for LSTM network but a mandatory

step of converting our words into tokens (integers)

Embedding Layer : that converts our word tokens (integers) into

embedding of specific size
LSTM Layer : defined by hidden state dims and number of layers
Fully Connected Layer : that maps output of LSTM layer to a desired
output size
Sigmoid Activation Layer : that turns all output values in a value
between 0 and 1
Output : Sigmoid output from the last timestep is considered as the
final output of this network
 Training LSTM for Sentiment Analysis
Training an LSTM model for sentiment analysis involves preprocessing text data,
embedding words into vectors, and finetuning the LSTM layers. The model learns to
extract sentiment related features from the input text.

Evaluation Metrics
When evaluating the performance of an LSTM-based sentiment analysis model, accuracy,
precision, recall, and F1 score are commonly used metrics. These metrics help gauge the
model's effectiveness in capturing sentiment.
 Future Research and
Developments
The field of sentiment analysis is continually evolving. Future research may
focus on multimodal sentiment analysis, context aware sentiment modeling,
and real-time sentiment prediction using LSTM and other advanced
techniques
We can extend this study to a larger extent where different embedding models
can be considered on large variety of the datasets.
Challenges and Opportunities

While LSTM offers significant advancements in sentiment analysis, there are

challenges such as model interpretability and handling noisy data.
Overcoming these challenges presents opportunities for further innovation in
the field.
Conclusion
In conclusion, leveraging LSTM for sentiment analysis enhances the accuracy and contextual
understanding of sentiment in textual data. As organizations strive to extract meaningful
insights from vast amounts of text, LSTM proves to be a powerful tool in sentiment analysis.
References
[1]Text-based Sentiment Analysis using LSTM
[2]Sentiment Analysis using Neural Network and LSTM
[3]Maas, A. L., Daly, R. E., Pham, P. T., Huang, D., Ng, A. Y., and Potts, C.
(2011, June). Learning word vectors for sentiment analysis. In
Proceedings of the 49th Annual Meeting of the Association for
Computational Linguistics: Human Language Technologies-Volume 1 (pp.
142-150). Association for Computational Linguistics
[4] ApoorvAgarwal, BoyiXie, Ilia Vovsha, Owen Rambow, Rebecca
Passonneau. Sentiment Analysis on twitter data. Department of Computer
Science, Columbia University, New York, NY 10027 USA.
Thank You