Controlling Output Length in Neural Encoder-Decoders

TL;DR

This paper introduces methods to control output length in neural encoder-decoder models, enabling more precise and concise sequence generation for applications like summarization without sacrificing quality.

Contribution

It proposes four novel methods, two decoding-based and two learning-based, for effectively controlling output length in neural sequence generation models.

Findings

Learning-based methods successfully control output length

Controlled outputs maintain summary quality

Methods outperform baseline in length accuracy

Abstract

Neural encoder-decoder models have shown great success in many sequence generation tasks. However, previous work has not investigated situations in which we would like to control the length of encoder-decoder outputs. This capability is crucial for applications such as text summarization, in which we have to generate concise summaries with a desired length. In this paper, we propose methods for controlling the output sequence length for neural encoder-decoder models: two decoding-based methods and two learning-based methods. Results show that our learning-based methods have the capability to control length without degrading summary quality in a summarization task.