Light RAT-SQL: A RAT-SQL with More Abstraction and Less Embedding of Pre-existing Relations

Abstract

RAT-SQL is among the popular framework used in the Text-To-SQL challenges for jointly encoding the database relations and questions in a way to improve the semantic parser. In this work, we propose a light version of the RAT-SQL where we dramatically reduced the number of the preexisting relations from 55 to 7 (Light RAT-SQL-7) while preserving the same parsing accuracy. To ensure the effectiveness of our approach, we trained a Light RAT-SQL-2, (with 2 embeddings) to show that there is a statistically significant difference between RAT-SQL and Light RAT-SQL-2 while Light RAT-SQL-7 can compete with RAT-SQL. Keywords: Deep learning, Natural Language Processing, Neural Semantic Parsing, Relation Aware Transformer, RAT-SQL, Text-To-SQL, Transformer.

Topological Relation Aware Transformer

“…In this work, our proposed method improves the encoder side, especially, the existing RAT-SQL model [8,7]. Please refer to [33,34,62] works for a thorough description of the decoder side.…”

“…In Text-To-SQL [1][2][3][4][5][6][7][8][9], The Light RAT-SQL [8] shows the way to reduce the number of preexisting relations in the RAT-SQL framework [7] by preserving the exact match accuracy without any enhancement of pre-trained (LLMs) large language models [10][11][12][13][14][15]. The limitation of this method is that it can not be suitable for the scenario where we have a lot of pre-existing relations.…”