LIU Yinuo

My research interests are mainly about language processing, and I hope to use a variety of technologies to explore the mysteries of human language and the brain. We hear people speak word by word, but we can build a syntactic tree with meanings related to world knowledge and concepts in a very quick time. I'm interested how the brain supports such a complex computational process. I want to combine cognivitve models and neural activities to answer questions related.

My future research will be conducted in two lines with questions including but not limited to:
(1) Bottom-up: How do we distinguish whether or not it is a speech sound? How do we encode and represent the sound? How do we combine sublexical input to derive a lexical meaning?
(2) Top-down: English words are ambiguous (e.g., one-to-many mapping, that the same word can have different meanings in different contexts). How the context helps in the ambiguity resolution? What is the temporal window? How are non-linguistic systems (e.g., memory regions) involved in the process?

I have done some language-related work in my past research training while honing my foundational skills in research on other cognitive funcitons (like working memory). Here are my major research projects.

Language

Lateralization of Language Brain Networks: A Graph Theory Study

In this research, I used graph theory to examine substantial lateralization in the topological properties of language brain networks and how they relate to behaviors. This work has been postered at OHBM 2022 Annual Meeting. I'm now examining the brain- behavior associations and the genetic heredity of the topological properties.

I learned how to do brain connectivity analysis using fMRI and some basic knowledge about graph theory. And my understanding of language brain networks was also deepened. I'm now working on the manuscript.

see: Slides , Poster

Interpretable and Table-structured Abstractive Text Summarization

This is my research in natural language processing in the Project-based Learning Program organized by MIT. In this research, I developed a text summarization algorithm with higher interpretability and more controllable output. It will first extract information (like entities, POS, etc.) and construct a structured table. After linearization, it will be imported to a fine-tuned T5-small model and get the final summaries.

This project made me an interdisciplinary student. I mastered the python language and become familiar with natural language processing by this program. It made me realize that language can be studied in many ways and aspects.

see: Manuscript

Working Memory

Posterior Activities During Encoding and Early Delay Support the Context Binding During Visual Working Memory

In this research, I explored a more effective behavioral paradigm to distinguish misbinding errors and informed guesses in the delayed recall paradigm by combining model fitting and individual reporting. I conducted EEG experiments and analyzed EEG signals (ERP, alpha power) to explore neural activity differences in different cognitive states to elucidate the neural activities closely related to feature binding in visual working memory. Preliminary results have been submitted to the CNS(Chinese Neuroscience Community) as a conference abstract.

This project made me think more about what scientific research is and how to scientific research in psychology. I also learned EEG signal acquisition and analysis. Although I may not continue to explore the topic of working memory in the future, this program enabled me to accumulate basic research skills.

see: Abstract

Spatial and Non-spatial Working Memory Recall Are Substantially Different: an Exploration Based on the Feature-binding Theory

In this work, I used the three-factor mixed response model (Bays, 2009) and the two-dimensional recall model (Grogan et al., 2019) to fit behavioral responses. I proved that the guessing rate is much lower in location recall tasks, and this has nothing to do with dots dimension, the way of presentation, and set size.

It can be called my introductory project for scientific research. It was a behavioral experiment in which I learned data collection, model fitting, and some basic data analysis methods. It was also the first time for me to make a poster, and the poster won the student Best Poster award in the scientific research achievements exhibition of our department, which encouraged me very much.

see: Poster

Modeling

A Simulation-Based Analysis of Fungi

I participated in the Mathematical Contest in Modeling (MCM/ICM) 2021 and this work won the Meritorious Winner Prize. In this work, I used the differential equation model and the cellular automata to simulate the interaction of different types of fungi.

This work improved my modeling skills and practiced my writing. It was the first time for me to participate in mathematical modeling. I was so excited to win the Meritorious Winner Prize.

see: Paper

Machine Learning Reveals Hemispheric Differences in the Human Brain

I helped with the Support Vector Machine (SVM) model and reviewed and edited the abstract paper. This work has a poster presentation in OHBM 2022 Annual Meeting, too.

This work enhanced my programming and modeling skills. I'm excited to combine artificial intelligence models with brain data to understand the features of human brain.

see: Slides