فرصة
Free participationCompetitions and Awards

Swahili News Classification NLP Learning Hackathon on Zindi

By: ZindiLast update:
Available

Applicant criteria

  • NationalityNo specific nationality required
  • AgeNo specific age required
  • GenderBoth

Opportunity criteria

  • Opportunity FieldForeign Languages and Literature, Others
  • LocationRemotely
  • Deadlineavailable

Opportunity description

Zindi is hosting the Swahili News Classification learning hackathon, a competition to build a model that sorts Swahili news articles into their specified categories using natural language processing. People from any country can participate individually or in teams of up to four, and the competition has no closing date.

About the Swahili News Classification Competition

The challenge is to develop a multi-class classification model that assigns each Swahili news item to its appropriate category. The intended use is to help Swahili online news platforms organize stories so readers can find the news they want. The project also aims to contribute to the continued representation of Swahili in apps and other online products.

Submissions are evaluated using Log Loss, with values ranging from zero to one. The source shows an example submission file with a test identifier and category columns including kitaifa, michezo, biashara, kimataifa and burudani. This is a learning competition, so there are no prizes. Participants receive 25 Zindi points for their first submission and 50 points for their first non-sample submission. Further information about Zindi points is available on the platform.

The competition does not close under its current timeline, although the organizers reserve the right to update the timeline if necessary. Participants may submit up to 10 times per day. A submission that fails during processing is not scored, does not count toward the daily limit and does not receive a score. There is no public and private leaderboard split.

Data, models and participation rules

Solutions must use publicly available, open-source packages, and models must not use the provided metadata. Participants may use only the datasets provided for this competition. Automated machine learning tools such as AutoML are not allowed. Pretrained models are permitted if they are openly available to everyone.

Competition data belongs to Zindi and the competition host. Participants must not transmit, duplicate, publish, redistribute or otherwise make the data available to anyone who is not participating. This includes uploading it to public sites such as Kaggle or GitHub. Cloud platforms such as Google Colab and AWS may be used if the data remains private and the use complies with Zindi rules. Participants must notify Zindi if they learn of unauthorized data transmission or access, and work with Zindi to address it.

Participation may be individual or in a team of up to four people. A team’s permitted submissions are calculated using the competition limit minus the highest submission count among its members when the team is formed. Multiple accounts per user and membership or collaboration across multiple teams are not permitted. Code cannot be privately shared outside a team; code that is shared must be made available to all participants through the platform, such as on the discussion boards. The rules also prohibit solutions that infringe third-party rights. Zindi may disqualify individuals, teams, accounts or submissions for rule violations, and the Terms of Use and FAQs may contain additional applicable rules. Explore other competitions and challenges, browse AI opportunities, or find more opportunities open to Arab applicants.

Who Should Apply

This challenge is suitable for people interested in natural language processing, text classification and building models for Swahili news content. It is open across countries, with no nationality restriction stated in the rules. Participants can work alone or with a small team, provided they follow the competition requirements for submissions, accounts, code and data. The source does not specify an academic qualification or a language proficiency requirement.

Key Highlights

  • Type: Learning competition focused on Swahili news classification.
  • Field: Natural language processing and text classification.
  • Participation: Individual or team of up to four.
  • Evaluation: Log Loss, with values from zero to one.
  • Deadline: No closing date under the published timeline.
  • Prizes: No prizes; Zindi points are awarded for the first and first non-sample submissions.

Benefits

  • Receive 25 Zindi points for the first submission.
  • Receive 50 Zindi points for the first non-sample submission.
  • Gain knowledge through a learning competition; the competition offers no prizes.

Eligibility criteria

  • The competition is open to everyone and is not restricted to any country.
  • Participants may compete individually or in teams of no more than four people.
  • Multiple accounts per user and membership or collaboration across multiple teams are not permitted.
  • Solutions must use publicly available, open-source packages, and models must not use the provided metadata.
  • Only the datasets provided for the competition may be used. Automated machine learning tools such as AutoML are prohibited. Pretrained models are allowed if openly available to everyone.
  • Competition data must not be made available to non-participants or uploaded to public sites. Cloud platforms may be used if the data remains private and Zindi rules are followed.
  • Code must not be privately shared outside a team; code that is shared must be made available to all participants through the platform.
  • Solutions must not infringe third-party rights. The maximum is 10 submissions per day.

Ready to apply?

Swahili News Classification NLP Learning Hackathon on Zindi