Final discussion
In a group of about three people, discuss each of the following.
You may also post comments to Oncourse after class.
Feel free to use the internet to help you out.
- For each of the following tasks, discuss
- what the components of the system would be
- what existing resources you would look for
- what additional resources you would need to develop, and how you might go about this.
- Your task is to design a translation interface that will allow deaf Indians to interact with
government offices.
The users will communicate using a dialect of Indo-Pakistani Sign Language (IPSL), and the government employees will use spoken or written Hindi (you decide).
(In later stages the project will incorporate other regional Indian languages.)
Like other sign languages, IPSL has all of the familiar grammatical and phonological properties of spoken languages, but these are of course realized in a spatial/visual form rather than an acoustic/auditory one.
Unlike some western sign languages like American Sign Language, however, there are very few resources for IPSL.
Basic resources such as part-of-speech taggers and large monolingual corpora do exist for Hindi.
- Your task is to design a system that reads news stories and determines the likelihood that they are factual, that the information reported in the article can be trusted.
- The field of computational linguistics/natural language processing has borrowed most of its
theories and methods from linguistics, AI (especially machine learning), and information retrieval.
How could the field benefit by learning more from psychology and cognitive science more generally?
- At the beginning of the semester, we read an article by Ted Pedersen critical of CL/NLP for its general failure to produce replicable research, for a sort of "faith-based empiricism".
Based on the papers we have read and, for some of you, your experience trying to implement the systems described in papers, how accurate would you say Pedersen's criticism is?