Representation Learning for Low-resource Language Processing

Date:

In this talk, I will present our works on representation learning for low-resource language processing.

Links: Presentation

======

Representation learning has emerged as an indispensable ingredient for low-resource language processing. Due to the differences across languages at levels of morphology, syntax, and semantics, it remains challenging to learn generalizable representations and effectively transfer them to a broad spectrum of languages. The first part of this talk will shed light on the difficulties of transferring representations across languages learned by two families of neural architectures (sequential RNN v.s. self-attention) due to word order differences. Furthermore, we present an approach to learn syntactic dependencies between words, facilitating cross-lingual information extraction. The second part of this talk will discuss our recent work on joint representing learning for programming and natural language that achieves state-of-the-art performance on various downstream software engineering tasks.