ALL-IN-ONE EMOTION, SENTIMENT AND

INTENSITY PREDICTION USING A

 Introduction

 Existing system

 Disadvantages

 Problem statement

 Proposed system

 Advantages

 System requirement specifications

1. S/W & H/W

2. Functional & Non-Functional

 SYSTEM DESIGN-ARCHITECTURE

 ALGORITHM

 .MODULES

 DATA FLOW DIAGRAM/UML DIAGRAMS

 Conclusion
ABSTRACT
We provide a multi-task ensemble framework that works on a number of connected tasks
at once. The ensemble model intends to use a hand-crafted feature representation for the
predictions along with three deep learning models' learned representations (i.e., CNN,
LSTM, and GRU). We address four issues related to emotion and sentiment analysis through
a multi-task framework, including "emotion classification & intensity," "valence, arousal &
dominance for emotion," "valence & arousal for sentiment," and "3-class categorical & 5-
class ordinal classification for sentiment." The fundamental issues span multiple domains
and two granularities (i.e., coarse- and fine-grained) (i.e., tweets, Facebook posts, news
headlines, blogs, letters etc.). According to experimental findings, the suggested multitask

framework performs better than the single-task frameworks across all experiments .
INTRODUCTION
 The automatic extraction of emotions from user-written texts is the subject of emotion
analysis. Ekman identified anger, contempt, fear, surprise, sadness, and joy as the
fundamental human emotions. As opposed to this, sentiment analysis attempts to
automatically extract the subjective data from user-written textual content and categorise
it into one of a preset range of categories, such as positive, negative, neutral, or conflict.
Sentiment and emotion are closely connected and frequently used improperly in the same
context. Munezero et al. claim that the scale of duration on which emotions and feelings
are perceived varies. While sentiments are more solid and valid for a longer amount of time
than emotions, which often have shorter durations. Additionally, emotions are not always
target-centric, whereas sentiments are typically directed towards a target entity. For
instance, someone might wake up joyful without any good reason.
EXISTING SYSTEM
 One of the earliest studies on emotion recognition classifies texts based on
the phrases that express different emotions. In a different study, Dung et
al. investigated how emotions affect people's mental states in order to train
a hidden markov model (HMM). The previous literature, in contrast,
proposed a rule-based method for extracting emotion-specific semantics,
which is subsequently applied for learning through various separable
mixture models. In contrast to the research above & focused primarily on
intensity prediction, these systems concentrated on emotion classification.
Jain et al. combined five distinct neural network models to predict the
strength of the emotion. In one of the models, they also looked at the
concept of multitask learning by treating four various emotions as the four
tasks. An average of the basis models, weighted, produced the final
forecasts.
DISADVANTAGES
 They demonstrated how higher-level activities can perform better when low-
level tasks are being watched over at lower levels of deep networks.
 Despite being a strong tool with many advantages, NLP nevertheless has a lot of
restrictions and issues, including the following: Words, phrases, and homonyms
used in context. Synonyms. Sarcasm and irony.)
PROBLEM STATEMENT
 In this study, four problems are simultaneously solved: coarse-grained
(categorical) emotion classification, fine-grained (valence, arousal, and
dominance) emotion prediction, fine-grained (valence and arousal)
sentiment prediction, and coarse-grained (categorical and ordinal)
sentiment classification. To do this, we offer a productive multi-task
ensemble architecture.
PROPOSED SYSTEM
 This is a major source of inspiration for the multi-task framework we've
proposed, which combines many tasks. Our framework's foundation is an
ensemble method. First, using three deep learning models—Convolutional
Neural Network (CNN), Long Short-Term Memory (LSTM), and Gated
Recurrent Unit—we learn hidden representations (GRU). Then, to create
an ensemble, we feed a Multi-Layer Perceptron (MLP) network with a
hand-crafted feature vector and the learned representations of three deep
learning systems. Utilizing four distinct representations, the goal is to
identify the crucial elements for predictions. The suggested network seeks
to simultaneously predict a number of outputs from input representations.
ADVANTAGES
 improved generality is attained
 enhances each task's performance through shared representation
 reduces the number of models needed from separate models for each
task in a single-task environment to only one unified model.
SYSTEM REQUIREMENT SPECIFICATIONS

Hardware Requirements
 1)Operating System : Windows Only
 2)Processor : i5 and above
 3)Ram : 4gb and above
 4)Hard Disk : 50 GB
Software Requirement
 1)Visual Studio Community Version
 2)Nodejs ( Version 12.3.1)
 3)Python IDEL ( Python 3.7 )
FUNCTIONAL REQUIREMENTS
Functional requirements are represented or stated in the form of input to be given to
the system, the operation performed and the output expected. System should
collect the data from any resources. All the collected data should be processed
for proper use, some analysis should be done for understanding the data
properly.
 Upload Social Media Dataset
 Data Preprocessing
 Feature Extraction
 Model Generation
 Upload Test Data
 Predict Result
NON-FUNCTIONAL REQUIREMENTS
 Usability
Usability is the main non-functional requirement for the “All-in-One Emotion, Sentiment and
Intensity Prediction Using a Multi-Task Ensemble Framework”. The UI should be simple
enough for everyone to understand and get the relevant information without any special
training. Different languages can be provided based on the requirements.
 Accuracy
Accuracy is another important non-functional requirement for the “All-in-One Emotion,
Sentiment and Intensity Prediction Using a Multi-Task Ensemble Framework”. The
dataset is used to Train and Test Model in python .Prediction should be correct, consistent,
and reliable.
 Availability
The System should be available for the duration when the user operates and must be recovered
within an hour or less if it fails. The system should respond to the requests within two
seconds or less.
 Maintainability
The software should be easily maintainable and adding new features and making changes to
the software must be as simple as possible. In addition to this, the software must also be
portable.
SYSTEM DESIGN ARCHITECTURE
ALGORITHMS
 CNN: A Convolutional Neural Network, also known as CNN or ConvNet, is a
class of neural networks that specializes in processing data that has a grid-like
topology, such as an image.
 LSTM: Long short-term memory (LSTM) is an artificial neural network used in
the fields of artificial intelligence and deep learning.
 GRU: Gated recurrent units (GRUs) are a gating mechanism in recurrent
neural networks. The GRU is like a long short-term memory (LSTM) with a
forget gate, but has fewer parameters than LSTM, as it lacks an output gate.
 MLP: Multi layer perceptron (MLP) is a supplement of feed forward neural
network. It consists of three types of layers—the input layer, output layer and
hidden layer, as shown in Fig. 3. The input layer receives the input signal to be
processed.
 BERT: BERT makes use of Transformer, an attention mechanism that learns
contextual relations between words (or sub-words) in a text.
ALGORITHMS
 BiLSTM: Bidirectional LSTM (BiLSTM) is a recurrent neural network used
primarily on natural language processing.
 RNN: A recurrent neural network (RNN) is a class of artificial neural networks
where connections between nodes can create a cycle, allowing output from
some nodes to affect subsequent input to the same nodes. 
 VC[CNN,LSTM,GRU],
 VC[CNN,LSTM,GAN,MLP] & VC[CNN,BiLSTM] Classifiers
MODULES
 Data exploration: using this module we will load data into system
 Processing: Using the module we will read data for processing
 Splitting data into train & test: using this module data will be divided into
train & test
 Model generation: Build
CNN,LSTM,GRU,MLP,BERT,BiLSTM,RNN,VC[CNN,LSTM,GRU],VC[CNN,LST
M,GAN,MLP] & VC[CNN,BiLSTM] Classifiers. Algorithms accuracy calculated.
 User signup & login: Using this module will get registration and login
 User input: Using this module will give input for prediction
 Prediction: final predicted displayed
DATAFLOW DIAGRAM
USECASE DIAGRAM
CLASS DIAGRAM
USECASE DIAGRAM
OBJECT DIAGRAM
STATE DIAGRAM
ACTIVITY DIAGRAM
SEQUENCE DIAGRAM
COLLABORATION DIAGRAM
COMPONENT DIAGRAM
DEPLOYMENT DIAGRAM
CONCLUSION
 In this study, we have put forth a multi-task ensemble architecture for
sentiment analysis, intensity prediction, and emotion analysis. We used an
MLP network, which collaboratively learns several related tasks, for the
ensemble. To extract the learned representations, we first created three
distinct deep learning models (i.e., CNN, LSTM, and GRU).
EXTENSION
 In this paper the author had applied CNN (accuracy is 80.52), LSTM (accuracy is
84), GRU (accuracy is 84), MLP (accuracy is 78), Ensemble of CNN, LSTM and
GRU (accuracy is 85), Ensemble of CNN, LSTM, GAN and MLP (accuracy is 89)
for multi-task learning. For Single-task learning, CNN (accuracy is 79), LSTM
(accuracy is 84), GRU (accuracy is 83), MLP (accuracy is 78), Ensemble of CNN,
LSTM and GRU (accuracy is 85), Ensemble of CNN, LSTM, GAN and MLP
(accuracy is 89).
 So as an extension to existing work we will apply some other techniques like
BERT model.
 As an extension we will implement some more deep learning techniques like
BiLSTM, RNN.
 As an extension we built ensemble model of CNN and BiLSTM.
 For this project we will implement Front end using Flask Framework as an
extension.
 We will provide user authentication using sqlite3 database.Base paper Link -
https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8756111