Learning Better Intent Representations for Financial Open Intent Classification

TL;DR

This paper enhances open intent classification in financial virtual agents by integrating prefix-tuning and last-layer fine-tuning of large language models, improving accuracy with minimal additional parameters.

Contribution

It introduces a novel combination of supervised pre-training methods with adaptive decision boundary post-processing, achieving state-of-the-art results with minimal parameter increase.

Findings

Achieved 1.63%-2.07% higher accuracy than prior methods.

Supplemented ADB with only 0.1% additional trainable parameters.

Outperformed full fine-tuning in ablation studies.

Abstract

With the recent surge of NLP technologies in the financial domain, banks and other financial entities have adopted virtual agents (VA) to assist customers. A challenging problem for VAs in this domain is determining a user's reason or intent for contacting the VA, especially when the intent was unseen or open during the VA's training. One method for handling open intents is adaptive decision boundary (ADB) post-processing, which learns tight decision boundaries from intent representations to separate known and open intents. We propose incorporating two methods for supervised pre-training of intent representations: prefix-tuning and fine-tuning just the last layer of a large language model (LLM). With this proposal, our accuracy is 1.63% - 2.07% higher than the prior state-of-the-art ADB method for open intent classification on the banking77 benchmark amongst others. Notably, we only supplement the original ADB model with 0.1% additional trainable parameters. Ablation studies also determine that our method yields better results than full fine-tuning the entire model. We hypothesize that our findings could stimulate a new optimal method of downstream tuning that combines parameter efficient tuning modules with fine-tuning a subset of the base model's layers.