A Bi-Encoder LSTM Model For Learning Unstructured Dialogs

TL;DR

This paper introduces a Bi-Encoder LSTM model for learning unstructured multi-turn dialogs, improving response selection accuracy in retrieval-based chatbots using large dialog datasets.

Contribution

The paper proposes a novel LSTM-based architecture for dialog response selection and demonstrates its effectiveness on the Ubuntu Dialog Corpus with improved accuracy over benchmarks.

Findings

Achieved higher Recall@1, @2, @5 accuracy than benchmark models.

Evaluated multiple similarity functions and hyper-parameters.

Validated the model's effectiveness on large-scale dialog data.

Abstract

Creating a data-driven model that is trained on a large dataset of unstructured dialogs is a crucial step in developing Retrieval-based Chatbot systems. This paper presents a Long Short Term Memory (LSTM) based architecture that learns unstructured multi-turn dialogs and provides results on the task of selecting the best response from a collection of given responses. Ubuntu Dialog Corpus Version 2 was used as the corpus for training. We show that our model achieves 0.8%, 1.0% and 0.3% higher accuracy for Recall@1, Recall@2 and Recall@5 respectively than the benchmark model. We also show results on experiments performed by using several similarity functions, model hyper-parameters and word embeddings on the proposed architecture