An Exploratory Evaluation of LLM-Assisted Rewriting of Moderate-Complexity Financial Sentences for DisCoCat-Based Sentiment Analysis

2026-08-07

AI Summary

This work addresses key limitations of the DisCoCat framework in financial sentiment analysis—namely its syntactic sensitivity, high simulation costs, and difficulty handling moderately complex sentences. To overcome these challenges, the authors propose a large language model (LLM)-driven, circuit-aware preprocessing method that employs controlled rewriting to compress such sentences into forms amenable to efficient DisCoCat processing while preserving sentiment semantics. For the first time, this approach integrates multiple LLMs (e.g., GPT-4.1-mini), tailored prompting strategies, and filtering mechanisms, substantially reducing quantum resource requirements: the strongest compression scheme cuts average qubit count and gate numbers by over 70%. Using GPT-4.1-mini with Prompt B, the method achieves an accuracy of 0.550 ± 0.035, outperforming the baseline of 0.521 ± 0.050, and further uncovers a negative correlation between training set size and model performance.

Abstract

Quantum natural language processing (QNLP) provides a grammar-aware framework for text modeling, and Distributional Compositional Categorical (DisCoCat) is one of its theoretically grounded formulations. Prior work on financial sentiment analysis has identified practical limitations of DisCoCat, including parser sensitivity, high simulation cost, and difficulty handling longer sentences. We study an LLM-assisted preprocessing workflow that uses controlled rewriting to compress, simplify, or decompose moderate-complexity financial sentiment sentences into parser-compatible, circuit-efficient variants while preserving sentiment-bearing meaning. We compare prompting strategies, language models, and filtering configurations with the low-complexity-only DisCoCat baseline of Stein et al. At the circuit level, the strongest compression variants reduce average qubit and gate counts by more than 70 percent relative to the raw moderate-complexity subset. Across repeated training runs, GPT-4.1-mini with Prompt B achieves the highest observed mean accuracy, $0.550 \pm 0.035$, compared with $0.521 \pm 0.050$ for the baseline. Larger training splits do not necessarily improve downstream performance; across evaluated configurations, training-split size has a moderately negative association with accuracy (Pearson $r=-0.446$). These results provide exploratory evidence that LLM-assisted rewriting can make some moderate-complexity inputs usable within the evaluated DisCoCat configuration, while highlighting prompt design, filtering, and circuit-aware preprocessing as considerations for more scalable QNLP-based financial sentiment analysis.

Problem

• DisCoCat
• financial sentiment analysis
• quantum natural language processing
• parser sensitivity
• circuit complexity

Innovation

• LLM-assisted rewriting
• DisCoCat
• quantum natural language processing
• financial sentiment analysis
• circuit-aware preprocessing