BioCopy: A Plug-And-Play Span Copy Mechanism in Seq2Seq Models

TL;DR

BioCopy introduces a span copy mechanism for seq2seq models that improves long span copying by using BIO tags to guide token selection, leading to better performance on generative tasks.

Contribution

The paper presents BioCopy, a novel plug-and-play span copy mechanism utilizing BIO tags to enhance copying of long spans in seq2seq models.

Findings

Outperforms baseline models on two generative tasks

Effectively copies long spans with BIO-guided masking

Improves token retention during sequence generation

Abstract

Copy mechanisms explicitly obtain unchanged tokens from the source (input) sequence to generate the target (output) sequence under the neural seq2seq framework. However, most of the existing copy mechanisms only consider single word copying from the source sentences, which results in losing essential tokens while copying long spans. In this work, we propose a plug-and-play architecture, namely BioCopy, to alleviate the problem aforementioned. Specifically, in the training stage, we construct a BIO tag for each token and train the original model with BIO tags jointly. In the inference stage, the model will firstly predict the BIO tag at each time step, then conduct different mask strategies based on the predicted BIO label to diminish the scope of the probability distributions over the vocabulary list. Experimental results on two separate generative tasks show that they all outperform the baseline models by adding our BioCopy to the original model structure.