Triples-to-isiXhosa (T2X): Addressing the Challenges of Low-Resource Agglutinative Data-to-Text Generation

TL;DR

This paper introduces T2X, a new dataset for isiXhosa data-to-text generation, along with a novel subword-based model and evaluation framework, highlighting the challenges of low-resource, agglutinative language modeling.

Contribution

It presents T2X dataset, a new architecture SSPG for agglutinative languages, and an evaluation framework, advancing low-resource language data-to-text research.

Findings

Pretrained language models underperform on T2X.

Fine-tuning translation models yields best results.

SSPG outperforms existing models for agglutinative languages.

Abstract

Most data-to-text datasets are for English, so the difficulties of modelling data-to-text for low-resource languages are largely unexplored. In this paper we tackle data-to-text for isiXhosa, which is low-resource and agglutinative. We introduce Triples-to-isiXhosa (T2X), a new dataset based on a subset of WebNLG, which presents a new linguistic context that shifts modelling demands to subword-driven techniques. We also develop an evaluation framework for T2X that measures how accurately generated text describes the data. This enables future users of T2X to go beyond surface-level metrics in evaluation. On the modelling side we explore two classes of methods - dedicated data-to-text models trained from scratch and pretrained language models (PLMs). We propose a new dedicated architecture aimed at agglutinative data-to-text, the Subword Segmental Pointer Generator (SSPG). It jointly learns to segment words and copy entities, and outperforms existing dedicated models for 2 agglutinative languages (isiXhosa and Finnish). We investigate pretrained solutions for T2X, which reveals that standard PLMs come up short. Fine-tuning machine translation models emerges as the best method overall. These findings underscore the distinct challenge presented by T2X: neither well-established data-to-text architectures nor customary pretrained methodologies prove optimal. We conclude with a qualitative analysis of generation errors and an ablation study.