About
I'm Hosein Rezaei, a third year PhD student at the University of York.
My research focuses on Natural Language Processing (NLP), with a particular interest in Grounded Language Learning.
I've also been a web developer since 2013, though the start of my journey with computers and programming goes back to around 2004.

In the world of NLP:
I am primarily interested in Grounded Language Learning, which is an approach to semantics that views the meaning of a word as a kind of agreement between speakers about how to use that word. This contrasts with traditional GOFAI approaches to semantics, where meaning is derived from expert-created lexicons like WordNet, and with distributional semantics, where meaning is inferred from patterns of word co-occurrence in large text corpora.
For example, when people begin using a word in a new way—like how “google” became a verb—a chatbot using Grounded Language Learning could potentially pick up this new meaning through interaction, without needing to be trained on massive amounts of text. With this long-term goal in mind, I focus on exploring and developing such learning capabilities within Large Language Models (LLMs). Currently, my project involves placing LLMs as agents in text-based games to study if and how they can learn language by grounding it in actions and perceptions (read more).
Resume
Summary
Blockchain developer
2021 - 2022
RedAcre, Remote
- Development of web services to provide cryptographic operations such as signing and publishing transactions for downstream blockchain applications.
- Contributing to several cryptographic libraries written with Rust to improve their functionality with regard to threshold signature scheme (TSS) using HD on the curves Secp256k1 and Curve25519 to support various cryptocoins that use ECDSA and EdDSA.
- Developing a fiat payment gateway in a microservice structure to support seamless integration with various payment service providers.
Education
Master of Science & Computer Engineering, Software
2016 - 2018
Isfahan University of Technology, Iran
GPA 16.42 out of 20. Best scores in Data Mining, Text mining, and Statistical Pattern Recognition courses. Six projects are done, one in Databases, one for Word Sense Disambiguation (WSD), and others for Text summarization.
Bachelor of Science & Computer Science
2004 - 2011
Payam-e Noor University of Shahreza, Isfahan, Iran
GPA 14.12 out of 20. Best scores in Data Structures and Algorithms, Algorithm Design and Analysis, General Mathemeatics, Linear Algebra, and Theory of Computattion.
Part-time distant education, simultaneous with serveral part-time jobs. It was a hell of time really!
Professional Experience
Research Collaborator
July 2017 - December 2017
IT Center of Isfahan University of Technology, Iran
- R&D about data integration from various resources of the university into a central data warehouse for facilitating reporting and data analysis for high-level decision-makers.
- R&D about automatic software and hardware inventory, to help IT managers in the center having an updated, overall, and yet detailed view of the latest software and hardware in use all over the university.
- Worked with: Pentaho Business Analysis platform, PostgreSQL, OCSInventory, Drupal, etc.
Research collaborator
2016 - 2017
NLP Institute of Shahid Beheshti University, Tehran, Iran
Senior Web Developer
2013 - 2016 & 2018 - 2021
Partotech.com, Isfahan, Iran
- Development of several enterprise web applications in the brand-monitoring business. Our mission was to extract text, image, video, etc from web, social networks (e.g. Telegram, Twitter, and Instagram), TV signals, and newspapers and to inform our customers as immediately as possible whenever any content related to them are published.
- Maintenance of Khabarfarsi.com -- the first of its kind news search engine(just for Persian so far)-- and its backend projects including web crawler, news storage, and image storage. The latter is implemented by me from ground up.
Publications
Large Language Models
Interactive Text Games: Lookahead Is All You Need! Introducing three variations of LLMs that predict not only the next immidiate token but also the second, third, ... up to K future tokens. Read more
Text Classification
Improving performance by incorporating structural information of the texts using Graph Neural Networks (GNNs) and graph representations like AMR.
Extractive Text Summarization
Features in extractive supervised single-document summarization: case of Persian news.View on LREV
Word Embeddings
Word Embeddings Are Capable of Capturing Rhythmic Similarity of Words. View on Arxiv
Contact
Location:
Department of Computer Science
University of York
York, YO10 5GH
United Kingdom
Email:
hosein.rezaei@york.ac.uk