National Research Council Canada (NRC) has just announced a new postdoctoral fellowship (PDF). The PDF is for two years, assuming good progress in the first year. There is a list of projects that candidates for the PDF can apply to. One of them, on NN modeling for First Nations languages, could be ideal members of the ACL just finishing, or having recently earned, their PhD. The successful candidate would work with the two NLP teams at NRC.
NLP researchers at NRC were just allocated significant funding to manage a big project on aboriginal languages. The PDF is not formally linked to this big project, but it means the successful candidate will be in close touch with people at NRC also working on these languages.
GENERAL ANNOUNCEMENT OF PDF
English: http://www.nrc-cnrc.gc.ca/eng/careers/programs/postdoctoral_fellowships/...
Français: http://www.nrc-cnrc.gc.ca/fra/carrieres/programmes/bourses_recherche_pos...
At the bottom of the English page, there’s a blue button marked “View the list of projects”. If one clicks on this, one sees a list of 21 projects. The one relevant to ACL members is called:
Neural language models and translation systems for aboriginals in Canada
Location: Ottawa, Ontario, Canada
In recent years, the aboriginal languages of Canada have not benefited from advanced Natural Language Processing (NLP) technologies such as speech recognition, grammar-checking, and machine translation (MT) (MT). The research goal of this postdoctoral fellowship is to help redress this by building language models (LMs) and MT systems for aboriginal languages. LMs are core building blocks for some NLP systems – e.g., grammar checkers and speech recognizers. In phase 1 of the project, the successful candidate will build an LM for at least one aboriginal language (Inuktitut, Mohawk or Cree). In phase 2, the candidate will build an MT system for translating this language to and from English. These outputs of the project will be transferred to Canadian organizations that build language tools for aboriginal communities. Both phases pose an interesting research challenge: these morphologically complex languages are not amenable to traditional word-based approaches to building LMs and MT systems. Therefore, the project will focus on neural net (NN) techniques that exploit subword models. The successful candidate will work closely with two research teams that have considerable expertise in the application of NNs to NLP tasks; he or she will be encouraged to publish articles on the work.
Education: PhD in computer science or linguistics, with a specialization in computational linguistics or quantitative linguistics. Experience with neural nets/deep learning or with the computational linguistics of morphologically complex languages – preferably both – is highly desirable. Demonstrable interest in language preservation and revitalization would also be an asset.