Enhancing Foundation Models in Transaction Understanding with LLM-based Sentence Embeddings
Xiran Fan, Zhimeng Jiang, Chin-Chia Michael Yeh, Yuzhong Chen, Yingtong Dou, Menghai Pan, Yan Zheng
TL;DR
This paper tackles semantic information loss in transaction understanding caused by index-based categorical representations in tabular data. It introduces a hybrid framework that precomputes LLM-generated sentence embeddings as semantic initializations for lightweight foundation models, enabling production-friendly deployment. Key contributions include a multi-source data fusion pipeline, an explicit one-word limitation prompt design, and comprehensive empirical validation across multiple LLM architectures, showing consistent improvements in transaction understanding tasks. The approach improves semantic richness without incurring runtime LLM inference costs, making it impactful for real-world financial systems.
Abstract
The ubiquity of payment networks generates vast transactional data encoding rich consumer and merchant behavioral patterns. Recent foundation models for transaction analysis process tabular data sequentially but rely on index-based representations for categorical merchant fields, causing substantial semantic information loss by converting rich textual data into discrete tokens. While Large Language Models (LLMs) can address this limitation through superior semantic understanding, their computational overhead challenges real-time financial deployment. We introduce a hybrid framework that uses LLM-generated embeddings as semantic initializations for lightweight transaction models, balancing interpretability with operational efficiency. Our approach employs multi-source data fusion to enrich merchant categorical fields and a one-word constraint principle for consistent embedding generation across LLM architectures. We systematically address data quality through noise filtering and context-aware enrichment. Experiments on large-scale transaction datasets demonstrate significant performance improvements across multiple transaction understanding tasks.