Research Groups: Friday, March 19

 10:00 AM - 11:30 AM: Cognitive Science of Language Group
1. Guest speaker: Juliette Millet (Université Paris Diderot-Paris 7): "Inductive biases, pretraining and fine-tuning jointly account for brain responses to speech."

Our ability to comprehend speech remains, to date, unrivaled by deep learning models. This feat could result from the brain's ability to fine-tune generic sound representations for speech-specific processes. To test this hypothesis, we compare i) five types of deep neural networks to ii) human brain responses elicited by spoken sentences and recorded in 102 Dutch subjects using functional Magnetic Resonance Imaging (fMRI). Each network was either trained on an acoustics scene classification, a speech-to-text task (based on Bengali, English, or Dutch), or not trained. The similarity between each model and the brain is assessed by correlating their respective activations after an optimal linear projection. The differences in brain-similarity across networks revealed three main results. First, speech representations in the brain can be accounted for by random deep networks. Second, learning to classify acoustic scenes leads deep nets to increase their brain similarity. Third, learning to process phonetically-related speech inputs (i.e., Dutch vs English) leads deep nets to reach higher levels of brain-similarity than learning to process phonetically-distant speech inputs (i.e. Dutch vs Bengali). Together, these results suggest that the human brain fine-tunes its heavily-trained auditory hierarchy to learn to process speech.

2. Julia Watson (MSc., Department of Computer Science), Jai Aggarwal (MSc., Department of Computer Science), and Anna Kapron-King (MSc., Department of Computer Science): "Come together: Integrating perspective taking and perspectival expressions."

Conversational interaction involves integrating the perspectives of multiple interlocutors with varying knowledge and beliefs. An issue that has received little attention in cognitive modeling of pragmatics is how speakers deal with the choice of words like come that are inherently perspectival. How do such lexical perspectival items fit into a speaker's overall integration of conversational perspective? We present new experimental results on production of perspectival words, in which speakers have varying degrees of certainty about their addressee's perspective. We show that the Multiple Perspectives Model closely fits the empirical data, lending support to the hypothesis that use of perspectival words can be naturally accommodated as a type of conversational perspective taking.

1:00 PM - 2:30 PM: Fieldwork Group
Ana Tona Messina (Ph.D.): "A POS tagger for Nahuatl."

Nahuatl is the most widely spoken indigenous language in Mexico; it enjoys the attention of many academics and scholars in the country and abroad; and it has an active community of native researchers and young advocators. Yet, there aren´t many language resources for Nahuatl speakers. In this talk, I will look at the practical issues that stand in the way of progress in this particular case, and I present the first part-of-speech tagger for Nahuatl (still under construction).