Training Heterogeneous Features in Sequence to Sequence Tasks: Latent Enhanced Multi-filter Seq2Seq Model

TL;DR

The paper introduces a latent-enhanced multi-filter seq2seq model (LEMS) that improves training on heterogeneous language data by analyzing and clustering input representations in a latent space, enhancing convergence and performance.

Contribution

It proposes a novel model that uses latent space transformation and clustering to better handle heterogeneous features in sequence-to-sequence tasks.

Findings

Improved semantic parsing accuracy on Geo-query dataset.

Enhanced machine translation quality on Multi30k English-French.

Demonstrated better convergence with diverse training data.

Abstract

In language processing, training data with extremely large variance may lead to difficulty in the language model's convergence. It is difficult for the network parameters to adapt sentences with largely varied semantics or grammatical structures. To resolve this problem, we introduce a model that concentrates the each of the heterogeneous features in the input sentences. Building upon the encoder-decoder architecture, we design a latent-enhanced multi-filter seq2seq model (LEMS) that analyzes the input representations by introducing a latent space transformation and clustering. The representations are extracted from the final hidden state of the encoder and lie in the latent space. A latent space transformation is applied for enhancing the quality of the representations. Thus the clustering algorithm can easily separate samples based on the features of these representations. Multiple filters are trained by the features from their corresponding clusters, and the heterogeneity of the training data can be resolved accordingly. We conduct two sets of comparative experiments on semantic parsing and machine translation, using the Geo-query dataset and Multi30k English-French to demonstrate the enhancement our model has made respectively.