University of Petroleum and

Energy Studies

INTERNSHIP-LOW LEVEL DESIGN

on
Sentiment Analysis for Social media

BACHELOR OF TECHNOLOGY COMPUTER SCIENCE

AIML

SUBMITTED BY:
NAME ROLL NO. EMAIL

GUIDED BY:
SME Name
 Table of Contents
1. Introduction
1.1. Scope of the document
1.2. Intended audience
1.3. System overview

2. Low Level System Design

2.1. Sequence Diagram
2.2. Navigation Flow/UI Implementation
2.3. Screen Validations, Defaults and Attributes
2.4. Client-Side Validation Implementation
2.5. Server-Side Validation Implementation
2.6. Components Design Implementation
2.7. Configurations/Settings
2.8. Interfaces to other components

3. Data Design
3.1. List of Key Schemas/Tables in database
3.2. Details of access levels on key tables in scope
3.3. Key design considerations in data design

4. Details of other frameworks being used

4.1. Session Management
4.2.Caching

5. Unit Testing

6. Key notes

7. Reference

UPES|INTERNSHIP - 2024 2
 1. Introduction

● Purpose of the document

● Importance of sentiment analysis in social media
● Overview of the system development

2. Scope of the Document

● What aspects of sentiment analysis will be covered?

● Limitations and exclusions

3. Intended Audience

● Target audience for the document

● Prerequisites for understanding the content

4. System Overview

● Architecture of the deep learning system

o Data preprocessing
o Feature extraction
o Model selection and training
o Evaluation metrics
● Technology stack used
● Overview of the sentiment classification process

5. Data Collection and Preprocessing

● Sources of social media data

● Data cleaning techniques
● Tokenization and vectorization of text data

UPES|INTERNSHIP - 2024 3
 6. Model Development

● Selection of deep learning models

● Fine-tuning parameters for optimal performance
● Handling imbalanced data (if applicable)

7. Training and Evaluation

● Training process
● Validation and testing procedures
● Metrics for evaluating model performance (accuracy, precision, recall, F1-score)

8. Deployment Considerations

● Integration with social media platforms (if applicable)

● Scalability and real-time processing challenges

9. Case Study: Sentiment Analysis on SASM

● Application of the developed system to real-world social media data

● Results and insights gained

10. Conclusion

● Summary of the project

● Achievements and challenges encountered
● Future directions for improvement

11. References

● Citations for related research papers

● Links to relevant resources and tools used

UPES|INTERNSHIP - 2024 4
 2. LOW LEVEL SYSTEM DIAGRAM

1.1. SEQUENCE DIAGRAM:

UPES|INTERNSHIP - 2024 5
 1.2. NAVIGATION FLOW/ UI IMPLEMENTATION:
⮚ The user interface is designed for ease of use and accessibility, featuring:
● Main Dashboard: Provides an overview of the system capabilities and quick
access to various functionalities.
● Image Upload Section: Allows users to upload url or file for image/ text.
● Processing Status Indicator: Displays the progress of image processing.
● Results Display: Shows the original and restored images side-by-side for
comparison.
● Navigation: Intuitive navigation between the dashboard, upload section, and
results display.

1.3. SCREEN VALIDATION, DEFAULTS AND ATTRIBUTES:

⮚ Screen validations are crucial for ensuring the integrity and quality of the
input images:
● File Type Validation: Ensures only supported formats (JPEG, PNG) are
accepted.
● File Size Validation: Restricts the file size to prevent excessive load on the
server.
● Image Resolution: Automatically resizes images to fit within predefined
resolution limits for optimal processing.

UPES|INTERNSHIP - 2024 6
UPES|INTERNSHIP - 2024 7
 1.4. CLIENT-SIDE VALIDATION IMPLEMENTATION:
⮚ Client-side validation is implemented using JavaScript and HTML5,
providing instant feedback:
● File Input Validation: Checks file type and size before submission.
● Real-time Feedback: Alerts users to invalid inputs, reducing server load and
improving user experience.
● Drag and Drop Interface: Enhances usability with modern file input
methods.

1.5. SERVER-SIDE VALIDATION IMPLEMENTATION:

⮚ Server-side validation complements client-side checks to ensure robust
validation:
● Integrity Check: Verifies the file's integrity to prevent corrupted files from
being processed.
● Format Verification: Ensures the file adheres to the supported formats and
specifications.
● Security Measures: Implements security checks to prevent malicious file
uploads.

UPES|INTERNSHIP - 2024 8
UPES|INTERNSHIP - 2024 9
 1.6. COMPONENTS DESIGN IMPLEMENTATION:
⮚ The system is modular, comprising several interdependent components:
● Image Pre-processing Module: Includes functions for resizing,
normalization, and augmentation to prepare images for neural network input.
● CNN Model for Image Restoration: A deep learning model trained to
recognize and correct common image degradation issues.
● Post-processing Module: Applies final enhancements such as sharpening,
color correction, and artifact removal to the restored images.
● Logging Module: Records processing activities and errors for monitoring and
debugging purposes.
● Configuration Module: Manages settings such as model parameters, file
paths, and environmental variables.

1.7. CONFIGURATION / SETTINGS:

⮚ Configurations are critical for the flexible operation of the system:
● Model Parameters: Learning rate, batch size, and number of epochs for
training.
● File Paths: Directories for input images, processed images, and logs.
● Environment Settings: GPU/CPU configurations, memory allocation, and
other system resources.

1.8. INTERFACES TO OTHER COMPONENTS:

⮚ The system interfaces seamlessly with various external components:
● TensorFlow: For neural network operations and model execution.
● Flask: Serves as the web framework for handling HTTP requests and serving
the web application.
● OpenCV: Utilized for image processing tasks such as resizing, normalization,
and augmentation.
● Database Systems: For storing user data, image metadata, and logs.

UPES|INTERNSHIP - 2024 10
 3. DATA DESIGN

LIST OF KEYS SCHEMAS / TABLES IN DATABASE:

⮚ The database design includes several key schemas and tables to manage data
efficiently:
● Users Table: Stores user information, authentication details, and user-specific
settings.
● Images Table: Contains metadata about uploaded and processed images,
including file paths, upload timestamps, and processing status.
● Logs Table: Records system activities, errors, and processing details for
monitoring and debugging.
● Sessions Table: Manages user sessions and states to ensure a consistent user
experience.

UPES|INTERNSHIP - 2024 11
 DETAILS OF ACCESS LEVELS ON THE KEY TABLES IN SCOPE:
⮚ Access levels are meticulously defined to ensure data security and integrity:
● Admin Access: Full control over all tables, allowing for data management
and system configuration.
● User Access: Restricted access to user-specific data, including their uploaded
images and processing results.
● Read-Only Access: Limited access for guests or external systems to view
publicly shared data without making modifications.

KEY DESIGN CONSIDERATIONS IN DATA DESIGN:

⮚ Several key considerations guide the data design process:
● Scalability: Ensuring the database can efficiently handle growing volumes of
data and increasing user numbers.
● Data Integrity: Maintaining consistent and accurate data through constraints,
validations, and regular audits.
● Performance Optimization: Optimizing queries and indexing strategies to
ensure fast data retrieval and processing.
● Security: Implementing robust access controls, encryption, and regular
backups to protect sensitive data.

4. DETAILS OF OTHER FRAMEWORKS BEING USED

SESSION MANAGEMENT:
⮚ Session management is a critical aspect of the system, ensuring a seamless
and secure user experience:
● Flask-Session: Utilized for managing user sessions, storing session data on
the server-side to maintain user state across multiple interactions.
● Session Persistence: Ensures that user sessions persist across browser restarts
and network disruptions.
● Security Measures: Implements secure session cookies and regular session
timeouts to prevent unauthorized access.

UPES|INTERNSHIP - 2024 12
 CACHING:
⮚ Caching mechanisms are employed to enhance system performance by
reducing redundant computations:
● Redis Cache: A high-performance in-memory data store used to cache
frequently accessed data and intermediate processing results.
● Query Caching: Caches database query results to minimize database load and
speed up data retrieval.
● Model Caching: Stores precomputed neural network models and weights to
expedite model loading and execution.

5. UNIT TESTING:

⮚ Unit testing is an integral part of the development process, ensuring the reliability
and correctness of the system:
● Test Framework: Pytest is used for writing and executing unit tests, providing a
robust and flexible testing environment.
● Test Coverage: Comprehensive test cases cover all major components, including
image pre-processing, neural network operations, and post-processing.
● Continuous Integration: Automated testing pipelines are set up using tools like
Jenkins or GitHub Actions to ensure continuous integration and immediate
feedback on code changes.
● Mocking and Stubbing: Utilized to simulate external dependencies and isolate
components for accurate testing.

UPES|INTERNSHIP - 2024 13
 6. KEY NOTES:
a) Lessons Learned:

● The importance of high-quality training data for achieving optimal neural network
performance.
● The challenges of tuning neural network parameters and the iterative process
required for model optimization.
● The need for robust validation mechanisms to ensure input data integrity and
system reliability.

b) Challenges Faced:

● Handling large image datasets and ensuring efficient processing without

overwhelming system resources.
● Balancing model complexity with processing time to achieve real-time
performance.
● Integrating various system components and ensuring seamless
communication between them.

c) Future Improvements:

● Extending support for additional image enhancement techniques, such as

super-resolution and colorization.
● Incorporating advanced neural network architectures to further improve
restoration quality.
● Enhancing the user interface with more interactive features and real-time
feedback.
● Scaling the system to handle higher volumes of users and images, ensuring
robustness and performance under load.

UPES|INTERNSHIP - 2024 14
 7.REFERENCES:

Books:
● Pang, B., & Lee, L. (2008). Opinion mining and sentiment analysis. Foundations and
Trends in Information Retrieval, 2(1-2), 1-135.
● Go, A., Huang, L., & Bhayani, R. (2009). Twitter sentiment classification using distant
supervision. CS224N Project Report, Stanford.
● Pak, A., & Paroubek, P. (2010). Twitter as a corpus for sentiment analysis and opinion
mining. LREC, 10, 1320-1326.
● Thelwall, M., Buckley, K., Paltoglou, G., Cai, D., & Kappas, A. (2010). Sentiment
strength detection in short informal text. Journal of the American Society for Information
Science and Technology, 61(12), 2544-2558.
● Kaggle Datasets: There are several datasets available on Kaggle related to sentiment
analysis of social media posts that can be used for research and development purposes.
● GitHub Repositories: Many open-source projects on GitHub provide implementations
and datasets for sentiment analysis of social media posts.

UPES|INTERNSHIP - 2024 15