Africa's Talking X Mozilla MCV Kiswahili Hackathon Series - Dar Es Salaam, Tanzania

Nov 3, 2023, 5:30 AM – Nov 11, 2023, 12:00 PM

Join us at the Africa's Talking X Mozilla MCV Kiswahili Hackathon Series and be part of building scalable solutions for Africa! Are you an innovator in ML and Language localization? Are you curious how you can implement voice technology in machine models to aid localization, especially in Kiswahili? Get your Ticket today!

3
RSVPs

APIsCommunityHackathonOpen source

About this event

Join us at the Africa's Talking X Mozilla MCV Kiswahili Hackathon Series on voice technology, and build scalable solutions ready for the African market!

How to get your Zindi Account: https://zindi.africa/onboarding


  • Is machine language and language localization a skill you possess or want to build on? 
  • Are you curious about how you can implement voice technology in machine language models to aid localization, especially in Kiswahili? 
Then this will be the hackathon series for you!


Join our transformative 4-part hackathon series, focused on harnessing the potential of voice technology. These sessions are designed to equip participants with vital skills for utilizing the Mozilla Common Voice (MCV) Kiswahili dataset effectively, fostering a community of problem-solving developers. Alongside, dive into the realm of machine learning as we delve into model training and implementation. This holistic approach ensures not only proficiency in voice tech but also hands-on experience in building intelligent systems. Don't miss this opportunity to become a part of the voice technology revolution and make a tangible impact on problem-solving using cutting-edge tools and techniques.

Below we delve into the details of each session, and explore how this comprehensive hackathon series will empower you in the exciting world of voice technology.


I. Session One


This session will see participants get an introduction to the field of (broadly) Artificial Intelligence and Machine Learning. This will include

  1. A definition and disambiguation of concepts; AI, ML, DS, supervised and unsupervised learning, an algorithm, a model,
  2. A definition of tasks in the field and examples of applications that make use of these tasks, e.g. Machine Translation, Sentiment Analysis, Speech to Text/Automatic Speech Recognition, Text to Speech. We will also discuss what the datasets for each of these tasks looks like, how they are built, and what language coverage is currently present, particularly for African languages.
  3. Building Responsible and Ethical AI systems(along with the dangers/harms made possible by these systems). This will include some practical examples of systems that have caused harm and how they could be built better as well as highlighting several responsible AI projects

II. Session Two


This session will narrow our focus down to the Speech-to-Text(STT) or Automatic Speech Recognition(ASR) ecosystem. We will:

  1. Define the task, how is a model trained in a supervised versus an unsupervised context, what about zero-shot learning where there is no data available, kinds of datasets(elicited speech versus read speech),
  2. Break down the composite component of an end-to-end speech recognition system (acoustic model, language model, lexicon)
  3. Look at end user applications where STT technology is used. Showcase the spread of languages, wrt low resource languages, and demo the performance of some enterprise models that claim coverage over thousands of languages
  4. Introduce Common Voice, do a platform walkthrough, discuss the type of dataset that this data collection process gives us, briefly describe the story of Kiswahili community building activities done by the community, showcase various characteristics of the dataset through an exploratory data analysis.
  5. Introduce the STT toolkit that we will use in the hackathon(Coqui STT or Nvidia NEMO), go through installation and setting up, define the metrics that are used to articulate/track the performance of our models and what it is they track(Character Error Rate, Word Error Rate, Phone Error Rate, loss), could cover some additional features of the toolkit, eg. data augmentation techniques(also a notebook).
  • Get started with Nvidia NeMo here.
  • Get started with Coqui SST here.

III. Session Three


A notebook will be made available and a brief introduction session will be facilitated where we ensure that each individual is properly set up on the platform (such as Kaggle) and can work through the entire notebook without any errors. The walk-through will include highlighting the hyper-parameters section, since this is where individual efforts should be focused to attempt to get the best model possible.

IV. Hackathon I Kick-off


This will be the official hackathon kick-off. This will be a model training hackathon. There will be a walk-through of how to make submissions to the hackathon once you do have a working solution. In this case, a winner will be determined by ranking the models based on the evaluation metrics.

V. Hackathon I Closing and Winners Announced


The winners of the first hackathon are awarded and proceed to the next session of the Africa's Talking X Mozilla MCV Kiswahili Hackathon Series.

VI. Session Four


This session will be an STT implementation training for the next hackathon where we invite individuals and teams to come through and make use of an API available for Kiswahili and implement it in various domains. 

VII. Hackathon II Kick-off


This will be the official hackathon II kick-off. This will be an implementation hackathon. The hackathon will build on the previous hackathon where there was a model developed by the participants for training, and a panel of judges will be constituted to analyse each of the submissions/pitches at the various events.

VIII. Hackathon II Closing


The second hackathon of the Africa's Talking X Mozilla MCV Kiswahili Hackathon Series is closed.

IX. Pitching of projects, judging and awarding for Hackathon II


The participants of the second hackathon take part in the pitching session and judging of the Africa's Talking X Mozilla MCV Kiswahili Hackathon Series. Winners of Hackathon II are announced.


Key Focus Areas:

  • API Integration
  • Speech Recognition
  • Text-to-Speech (TTS) Applications
  • Voice-Enabled Applications
  • Voice Data Collection and Annotation
  • Multilingual and Cross-Cultural Voice Tech
  • Voice Analytics
  • Natural Language Processing (NLP) and Machine Learning Models

Target Audience:

  • Developers
  • Data Scientists
  • AI and ML Enthusiasts
  • Tech Entrepreneurs
  • Linguists and Language Enthusiasts
  • Educators and Students

What should you expect as an Attendee:

  • Access to Cutting-Edge Tools: Leverage Africa's Talking powerful APIs and SDKs, and Mozilla Common Voice project tools designed to make implementation seamless and hassle-free.
  • Apprenticeship & Support: Work closely with seasoned experts from Africa's Talking and Mozilla who will guide and support you throughout the hackathon as we build with you.
  • Networking Opportunities: Join as a member and connect with like-minded individuals, potential collaborators, and industry professionals.
  • Cash Prizes, Recognition and Qualify to Join our Marketplace Program: Stand a chance to win exciting prizes, gain recognition for your innovative solutions and
  • Impact Africa: Contribute to the technological and economic growth and development of Africa by providing solutions that simplify and enhance machine learning and localization processes.


New to Africa's Talking? Get started here.

Find the link to the Notion Page with Hackathon Details Here.

Developer Documentation (We welcome feedback)

THEME: Mozilla Connected Voice Kiswahili Hackathon Series

The winning solution will receive cash prizes as follows (For each Hackathon):


  1. Winner - $1,000 (Join Marketplace Program)
  2. 1st Runners Up - $500
  3. 2nd Runners Up - $200

More exclusive prizes and giveaways to attendees.

Contribute to open-source.

Elarian: Build Responsibly, Protect Customer Data.

Elarian is a customer engagement framework (Customer Data Platform) that helps you build robust applications with the customer as the unit of abstraction. Elarian features robust capability in handling end to end of customer data.

Africa's Talking powers communications solutions across Africa with simplified access to telco infrastructure, developers use our powerful SMS, USSD, Voice, Airtime, and Payments APIs to bring their ideas to life, as they build and sustain scalable businesses.

Agenda:

1. Warm-Up

  • Introductions
  • Form teams where need be.
  • Feedback
  • Feature requests

2. Sessions I - III

Note: This section needs additional setup.

The sessions will involve training and learning of the new technologies. 

3. Hackathon I Kick-off

Teams or individuals begin the model training hackathon.

4. Hackathon I Closing and Winers Announced

The model training hackathon is closed off and the winners announced.

5. Session IV

This session will be an STT implementation training for the next hackathon.

6. Hackathon II Kick-off

This will be the official hackathon II kick-off.

7. Hackathon II Closing

The second hackathon of the Africa's Talking X Mozilla MCV Kiswahili Hackathon Series is closed.

8. Pitching of projects, judging and awarding for Hackathon II

The participants of the second hackathon take part in pitching and judging, then winners of Hackathon II are announced.

Note: Carry a laptop

Opportunities:

  • Network
  • Innovate
  • Access to our Talent bench (You can get a job or a gig on the spot)

Submit your Idea:

Hackathon idea submission form

Gigs:

We would love to reach out to you so that you can build for our customers, please fill out this form with details to ensure we have your details:

  • GIG/HACK DEVELOPER PORTFOLIO FORM

Join community channels:

Slack:

Follow us on Twitter handles too:

  • @AfricasTalking
  • @ATCommunityNBO
  • @ElarianHQ

You can get our videos, recaps, and event interviews on our youtube channels, subscribe to get updates:

Africa's Talking community allows developers to learn skills for the modern-day African Developer. We are language and framework agnostic. All developers are welcome. This is where Africa's Talking developers community meets to build, learn and exchange knowledge.

We are helping software developers and businesses to bring their ideas to life through easy-to-use APIs quickly.

Would you like to partner with us? Kindly contact the Developer Experience Team.

Speakers

  • Kalebu Gwalugano Jordan

    Neuro Tech

    CEO

  • Sylvia Jebet Kipkemoi

    Elarian

    Developer Relations

  • Mainye Ben

    Africa's Talking LTD

    Data Scientist & Maker

  • Michael Kimathi

    Africa's Talking

    Head of Developer Community Africa

  • Steve Waweru

    Tech Innovators Network

    Tech Lead

Facilitators

  • Veronica Michael

  • Bennet Benedict

When

When

November 3 – 11, 2023
5:30 AM – 12:00 PM UTC

Agenda

Warm-Up
Pitching of Solutions
Lunch
Pitching Continues
Winners Announcement

Hosts

  • Veronica Michael

  • Bennet Benedict

  • Graham Ingokho Muhanga

    Africa's Talking LLC

    SVP Marketing

  • Michael Kimathi

    Africa's Talking

    Head of Developer Community Africa

  • George Machibya

    Africastalking

    Developer Relations TZ

Organizers

  • Bennett Benedict

    University of Dar Es Salaam

    Tanzania, Community Lead,

  • Veronica Michael

    Tanzania, Community Lead

  • Jessica Randall

    University of the Western Cape

    South Africa Community Lead

  • Josphat Mwangi

    Kenya Community Lead

  • Sheriffo Jarju

    Community Lead Gambia

  • Isophel Natwijuka

    Uganda Community Lead

  • Sanusi Abdulkadir

    Nigeria Kaduna Community Lead

  • Avuzwa Ntshongwana

    South Africa Community Co-Lead

  • Josh Venceslas

    Altech Group

    Co-Lead: Rwanda Community