US 2026/0065064 A1Published application

End-to-end task-oriented dialogue system with few-shot learning

AssigneeViettel Group

InventorsKHAC HOAI NAM BUI, QUANG CHIEU NGUYEN, THANH DO NGUYEN

Life of this patent

Estimating the term…

Loading patent…

Abstract

The invention proposes a novel method for developing a TOD model by representing input data as function calls and utilizing LLM as the foundational model to train multi-tasks via finetuning LLM with instruction. Furthermore, to implement the proposed model on LLMs with moderate sizes (fewer than 10 billion parameters), the invention also presents a finetuning LLM method to enhance the capability of these LLMs in terms of handling function-calling tasks. Finally, an effective training strategy with customized loss functions for each specific task is presented to optimize the training process. Experimental results on the standard MultiWOZ 2.2 dataset demonstrate the superior performance of the proposed method compared to existing approaches in this field of research.

Claims1

Dependent claims

A method of developing an end-to-end TOD system with few-shot learning comprises the following steps:

step 1:
proposing an LLM-powered end-to-end TOD pipeline;
with an architecture that comprises three main task components, that standardize the traditional TOD tasks such as NLU, DST, and NLG into a function-calling paradigm, which are function selection, function completion, and generate response, respectively;
step 2:
fine-tuning instruction LLM for multi-task learning of function-related tasks;
enhancing function-related tasks for LLMs using dialogues with instruction of function calling task;
step 3:
proposing efficient training strategies, including input data normalization and a design of task-specific loss functions for a training process;
an overall loss function for an entire model training process is measured by a sum of task-specific loss functions, including function selection, function completion, and generation response, respectively.

6 more elements

Dependencies are inferred from the claim text.