One of the main purposes of language is to communicate meaning. When hearing or reading an utterance we extract the underlying message without much effort. To achieve this, many different types of information must be integrated with each other. For example, for sentences with ambiguous syntactic structure, the human brain excels at using word semantics in combination with general knowledge of the world to understand the meaning of a sentence.
For my PhD project, I am investigating neural representations of semantic and syntactic information and their interaction during sentence processing. By using specific models like corpus-based word embeddings as an approximation of semantic content, I try to isolate word level semantics from other processes. Specifically, I apply multivariate decoding analyses to Magnetoencephalography (MEG) data to extract those spatio-temporal patterns within the neural signal that best relates to different features in the stimulus.