PhD - Multilingual Spoken Language Understanding for Low-Resourced Sub-Saharan African Languages F/M

About the Role

Your role is to carry out PhD work on: Automatic multilingual speech understanding for sub-Saharan African languages.

Context
Orange operates in 14 countries in Sub-Saharan Africa, focusing on data and artificial intelligence to enhance customer experience and establish itself as a leading digital operator in the region. Despite Africa's dynamic growth, it faces significant challenges, including a literacy rate of only 63% in 2023 in Sub-Saharan African countries, with many literate individuals lacking proficiency in French or English. This considerably limits access to digital technologies for a large portion of the population.

To address PhD thesis challenges, Orange aims to develop speech technologies to better understand customers who will be able to speak in their native languages. However, most existing voice-based solutions are available only in major international languages, making it difficult to cater to the continent's approximately 2,000 languages and dialects. The proposed research will explore innovative machine learning strategies to create speech recognition and speech understanding models for the languages spoken in the Orange footprint, utilizing end-to-end approaches that enhance efficiency and robustness.

Given the limited availability of textual data for African languages, the candidate will need to devise methods to overcome this challenge, focusing on speech analysis, cross-language intrinsic characteristics discovery and pooling, to minimize reliance on written annotations. The research will involve exploring neural network techniques, such as self-supervised learning and transfer learning, to identify meaningful acoustic units without human annotations.

About You

Required Diploma:

• Master 2 in Natural Language Processing, Computer Science, Data Science or Mathematics

Expected Proficiencies:

• Skills in Natural Language Processing (Automatic Speech Recognition, Spoken Language Understanding, Data Mining, GenAI or LLM)
• Strong skills in Deep Learning (architectures, algorithms and methods)
• Advanced knowledge of the Python language and deep learning libraries (huggingface_hub, etc.)
• In-depth knowledge of a deep learning framework (Pytorch, Tensorflow2, Jax, etc.)