SIGMA: A Dataset for Text-to-Code Semantic Parsing with Statistical Analysis

2023-12-15

International Conference on Machine Learning and Applications

AI Summary

Existing semantic parsing approaches (e.g., Text-to-SQL) are constrained by the limited expressiveness of formal query languages like SQL, hindering support for multi-perspective statistical analysis. Method: This paper introduces SIGMA—the first Text-to-Python semantic parsing dataset tailored for statistical analysis—comprising 6,000 natural language questions paired with executable Python code, covering four data retrieval patterns and forty statistical operation types. SIGMA reframes semantic parsing as executable statistical code generation, transcending the expressivity boundaries of traditional formal languages. Contribution/Results: We evaluate state-of-the-art models—including LGESQL, 5mBoP, and SLSQL—integrated with ELECTRA, T5, and GraPPa pretraining frameworks in end-to-end generation settings. LGESQL+ELECTRA achieves 83.37% structural accuracy, while 5mBoP+GraPPa+T5 attains 76.38% execution accuracy. These results demonstrate the feasibility—and highlight key challenges—of statistical-aware code generation via semantic parsing.

Technology Category

• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Machine Learning: Statistical Relational/Logic Learning
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories

Application Category

• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

In the semantic parsing domain, significant progress has been achieved in Text-to-SQL and question-answering tasks, both focused on extracting information from data sources in their native format. However, the inherent constraints of their formal meaning representations, such as SQL programming language, hinder their ability to analyze data from various perspectives, such as conducting statistical analyses. To address this limitation and inspire research in this field, we design SIGMA, a new dataset for Text-to-Code semantic parsing with statistical analysis. SIGMA consists of 6000 questions with corresponding Python code labels. The Python code labels in our dataset cover 4 types of query patterns, which return data in their original format, and 40 types of statistical analysis patterns, which perform statistical operations on the data. We evaluated the SIGMA dataset using three different baseline models: LGESQL, 5mBoP, and SLSQL. The experimental results show that the LGESQL model with ELECTRA outperforms all other models, achieving 83.37% structure accuracy. In terms of execution accuracy, the 5mBoP model, when combined with GraPPa and T5, reaches 76.38%.

Problem

• Addressing limitations in semantic parsing for statistical analysis
• Introducing SIGMA dataset for Text-to-Code with Python
• Evaluating models for accuracy in structure and execution

Innovation

• SIGMA dataset for Text-to-Code semantic parsing
• Python code labels for statistical analysis
• Evaluated with LGESQL, SmBoP, SLSQL models