translators without borders

 

patrick-tomasso-71909-unsplash.jpg
TWB logo.png

Help bridge the language gap in humanitarian work.

Improving humanitarian assistance by prioritising translation needs through automatic impact assessment.

Problem: When humanitarians respond to disasters, they need to be able to communicate with affected communities, NGOs, and locals on the ground. Often, however, they do not speak the same language. Volunteers from Translators Without Borders (TWB) are therefore absolutely crucial to providing humanitarian assistance.

TWB’s vast volunteer community translates large amounts of documents on behalf of partner NGOs. However, TWB knows little about the potential impact of the translated documents, and therefore cannot prioritise which documents to translate. This information would be essential in order to maximise the impact of our own work. Impact is defined here as a mix of factors such as document type, audience, sectors (Health, Logistics, Nutrition, Protection, Shelter, Water/Sanitation/Hygiene, Camp Coordination and Camp Management, Early Recovery, Education, Emergency Telecommunications, Food Security) and relevance for concurrent crises.   

Outcome: A solution that can:

a) Automatically assess the impact of translations produced by TWB in the past, and

b) Predict the impact of a new translation being requested.

For:

a) A static solution for a (repeatable) one-time analysis would suffice;

b) However, would ideally be available via an API for integration with TWB’s translation management system.

Datasets:

●     TWB corpus of ~5000 documents with the following metadata for each document:

○     original file format (.docx, .pdf, .pptx etc.)

○     time of translation request

○     requesting NGO

○     target language (source language is English for the vast majority of the documents)

○     target country (probably of limited trustworthiness)

The documents will be provided as text files.

●     Data from the United Nations Office for the Coordination of Humanitarian Affairs (OCHA)