Interpreting unknown words in context using structured sparse coding
Speaker: András Lörincz, Senior researcher, Eötvös Loránd University, Budapest, Hungary.
We propose a method based on structured sparse coding to interpret unknown words. We distinguish two types of unknown words: words whose surface form is corrupted, but the concept they denote is known, and novel words that are completely unknown. Novel words are interpreted as a combination of related concepts by exploiting the distributional hypothesis, according to which words that occur in the same contexts have similar meanings...