# Natural Language Generation for Spoken Dialogue System using RNN Encoder-Decoder Networks

## TL;DR

This paper introduces an RNN encoder-decoder model with attention and LSTM decoders for natural language generation in spoken dialogue systems, demonstrating superior performance and generalization across multiple domains.

## Contribution

It presents a novel RNN encoder-decoder architecture with attention and LSTM decoding for joint sentence planning and surface realization in NLG.

## Key findings

- Outperforms previous NLG methods across four datasets
- Shows strong generalization to unseen domains
- Learns effectively from multi-domain datasets

## Abstract

Natural language generation (NLG) is a critical component in a spoken dialogue system. This paper presents a Recurrent Neural Network based Encoder-Decoder architecture, in which an LSTM-based decoder is introduced to select, aggregate semantic elements produced by an attention mechanism over the input elements, and to produce the required utterances. The proposed generator can be jointly trained both sentence planning and surface realization to produce natural language sentences. The proposed model was extensively evaluated on four different NLG datasets. The experimental results showed that the proposed generators not only consistently outperform the previous methods across all the NLG domains but also show an ability to generalize from a new, unseen domain and learn from multi-domain datasets.