Grammar Assistance Using Syntactic Structures (GAUSS)
Project Overview
The document focuses on the GAUSS project, an innovative initiative designed to enhance grammar coaching for Spanish through a novel parsing algorithm that effectively merges linguistic formalism with practical application. It aims to overcome the shortcomings of current grammar coaching systems by offering meaningful feedback while minimizing the computational costs typically linked to neural approaches. The project emphasizes inclusivity in language education, thereby striving to improve grammar coaching tools not only for widely spoken languages like Spanish but also for minority languages. This highlights the potential of generative AI in creating adaptable and efficient educational technologies that can cater to diverse language learners, ultimately fostering a more inclusive and effective learning environment. The findings suggest that such advancements in AI-driven educational tools can significantly enhance the quality of language learning experiences, making them more accessible and responsive to individual needs.
Key Applications
Grammar Assistance Using Syntactic Structures (GAUSS)
Context: Developing a grammar coaching system for Spanish, targeting Spanish language learners.
Implementation: Utilizes a rich linguistic formalism and a fast parsing algorithm to provide feedback on writing.
Outcomes: Improves grammar coaching by giving meaningful feedback on grammatical structures and addressing biases in language usage.
Challenges: Integrating neural and symbolic approaches to parsing and addressing the speed bottleneck of HPSG parsing.
Implementation Barriers
Technical Barrier
The speed of HPSG parsing is slow due to the large feature structures leading to a huge search space. Existing grammar coaching systems often rely on expensive neural methods that require large quantities of training data.
Proposed Solutions: Improving analyses in the grammar to reduce ambiguity, integrating top-down parsing, filtering lexical entries, and developing a system less reliant on neural methods and more focused on linguistic formalism.
Project Team
Olga Zamaraeva
Researcher
Lorena S. Allegue
Researcher
Carlos Gómez-Rodríguez
Researcher
Anastasiia Ogneva
Researcher
Margarita Alonso-Ramos
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Olga Zamaraeva, Lorena S. Allegue, Carlos Gómez-Rodríguez, Anastasiia Ogneva, Margarita Alonso-Ramos
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai