SPRINT II
Objectives with User Stories of Sprint II
The main objective of this sprint is to enhance the Tamil Speech-to-Sign Language Translator
by expanding its vocabulary and improving its overall accuracy. The focus is on refining the
gesture mapping system to support more complex Tamil phrases, better accommodating
different accents and dialects, and introducing a feedback loop to drive continuous
improvements based on real-world usage.
User Stories
As a developer, I want to expand the gesture mapping system to include a broader
range of complex Tamil phrases and sentences, allowing the system to translate more
diverse and meaningful communication.
As a teacher, I want the system to recognize and accommodate the many Tamil
dialects and accents spoken across different regions, ensuring that it remains
accessible and effective for all my students.
As a student, I want the system to display sign language gestures for more advanced
academic terms, so that I can better understand lessons and participate actively in the
classroom.
As a researcher, I want to collect regular feedback from users to help refine the
system’s accuracy over time, ensuring it continues to evolve and meet the changing
needs of both students and educators.
Functional Document
1. Introduction
Building on the foundation laid during Sprint 1, this sprint is all about expanding the system’s
vocabulary and enhancing the accuracy of gesture mapping. We will integrate advanced
natural language processing (NLP) techniques to better handle complex Tamil phrases and
introduce systems that adapt to different regional dialects. Additionally, a feedback
mechanism will be set up to allow users to easily share their experiences, helping guide the
system’s continuous improvement.
2. Product Goal
By the end of this sprint, the system will be capable of:
Translating complex Tamil sentences, including academic terms, into accurate sign
language gestures.
Recognizing Tamil speech accurately even when spoken with different regional
accents and dialects.
Gathering and using user feedback to fuel continuous updates and improvements to
the system’s performance.
3. Demography
The enhanced system is primarily intended for educators and hearing-impaired students in
Tamil-speaking regions, creating a more inclusive and supportive learning environment. It
also aims to assist researchers who are working in the fields of accessibility and speech-to-
sign language translation. Additionally, DevOps and technical teams play a crucial role in
ensuring the system’s ongoing development, stability, and efficiency.
4. Business Processes
Data Collection: Gather Tamil speech samples that represent a wide range of accents
and dialects, along with regular feedback from real users.
Pre-processing: Further refine the text cleaning and tokenization processes to better
handle complex phrases and broader vocabulary sets.
Gesture Mapping: Expand the gesture database to include academic vocabulary,
greetings, and common classroom phrases.
Dialect Adaptation: Implement advanced methods to increase recognition accuracy
across the various dialects spoken in Tamil Nadu.
Real-Time Translation: Maintain a high standard of real-time performance, even when
processing longer and more complex sentences.
Evaluation: Continuously assess system performance and accuracy through real-world
use cases, leveraging user feedback for iterative improvements.
5. Features
F1 - Expanded Vocabulary: The system will now support complex and academic
Tamil phrases, enabling richer and more precise communication.
F2 - Regional Accent Support: Enhanced recognition capabilities will ensure the
system understands a diverse range of Tamil dialects.
F3 - Feedback Collection: Users will be able to report issues and suggestions easily,
helping guide ongoing refinement and development.
F4 - Enhanced Gesture Mapping: New gestures will be added to cover a broader set of
academic and classroom-related terms, making the system even more valuable for
educational settings.
[Link] Matrix
Role Access Level
Admin / Developer Full edit rights, model retraining, dataset management
Teacher Execute translation process, provide feedback, view sign
language gestures
Student View translated sign language gestures, submit feedback
for improvement
Read-only access to system performance metrics, logs, and
Researcher user feedback
7. Assumptions
The system will increase its gesture vocabulary by adding a larger set of Tamil Sign
Language gestures, not only for day-to-day communication but also for academic
terms and frequent greetings, enhancing interactions to be more complete and
meaningful.
The training and improvement dataset consists of Tamil speech examples from a large
variety of regional accents, such that the system becomes increasingly accurate and
efficient in the diverse linguistic environment of Tamil Nadu.
User feedback will be gathered regularly to constantly refine and improve the
accuracy of the gesture mapping process, making sure the system develops to meet
better the demands of the real world.
In order to ensure user privacy and trust, all speech data will be anonymized in full,
with no personally identifiable information (PII) stored or retained anywhere.
Outcome of Objectives / Result Analysis
At the completion of this sprint, the Tamil Speech-to-Sign Language Translator demonstrated
enhanced proficiency in handling more intricate Tamil sentences and phrases with greater
precision, particularly excelling in recognizing regional dialects. The gesture-mapping system
was also expanded to cover academic vocabulary, greetings, and common classroom
expressions.
Results:
Speech Recognition Accuracy:
Achieved an 85% accuracy rate for standard Tamil phrases, with notable
improvements in the comprehension of regional dialects.
Gesture Mapping Precision:
Reached 90% precision when translating academic and complex sentences. Minor
errors were identified and corrected through user feedback.
Real-Time Translation:
Successfully translated intricate sentences in real time, maintaining an average
response time of less than 3 seconds.
Feedback Integration:
80% of the feedback collected from students and educators led to meaningful system
updates, including the addition of new gestures and improved pronunciation
recognition for specific phrases.
Sprint Retrospective
What Went Well What Went Poorly Idea for Improvement Action Plan
Vocabulary expanded with Research and integrate models that
Some regional accents not recognized, leading to Implement advanced speech models for diverse
complex Tamil phrases and improve recognition across Tamil
translation errors. Tamil accents.
academic terms. dialects.
Gesture mapping was enhanced Refine mapping with context-based
Mismatches in gesture mapping for complex Improve gesture mapping for multi-word and
for academic and classroom checks to ensure accurate gesture
phrases. academic phrases.
terms. matching.
Automate feedback collection and
Real-time translation maintained Streamline the feedback process for quicker
Feedback loop was slow, delaying updates. processing for faster integration of user
with enhanced gesture mapping. updates.
suggestions.
Diverse Tamil speech samples Revamp feedback system with
Increase engagement by improving feedback
collected for better dialect Low user engagement with feedback system. notifications and incentives to
system usability.
support. encourage consistent user input.
