# Structure Before Sampling: Community-Aware Core-Set Selection for Data-Efficient Text-to-Speech

arXiv: https://arxiv.org/abs/2609.24275 · HTML: https://mlsanity.com/paper/2609.24275/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 10 of its 26 references that are in the map.
- First posted on arXiv 21 Sep 2026.

## Builds on (most cited first, 10 of 10)

- [Common Voice: A Massively-Multilingual Speech Corpus](https://mlsanity.com/paper/1912.06670/) (2019)
- [Deduplicating Training Data Makes Language Models Better](https://mlsanity.com/paper/2107.06499/) (2021)
- [XLS-R: Self-supervised Cross-lingual Speech Representation Learning at Scale](https://mlsanity.com/paper/2111.09296/) (2021)
- [Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech](https://mlsanity.com/paper/2106.06103/) (2021)
- [Beyond neural scaling laws: beating power law scaling via data pruning](https://mlsanity.com/paper/2206.14486/) (2022)
- [Unsupervised Cross-lingual Representation Learning for Speech Recognition](https://mlsanity.com/paper/2006.13979/) (2020)
- [SemDeDup: Data-efficient learning at web-scale through semantic deduplication](https://mlsanity.com/paper/2303.09540/) (2023)
- [Coverage-centric Coreset Selection for High Pruning Rates](https://mlsanity.com/paper/2210.15809/) (2022)
- [D2 Pruning: Message Passing for Balancing Diversity and Difficulty in Data Pruning](https://mlsanity.com/paper/2310.07931/) (2023)
- [Diversity-based core-set selection for text-to-speech with linguistic and acoustic features](https://mlsanity.com/paper/2309.08127/) (2023)

## Abstract (arXiv)

Text-to-speech (TTS) corpora are costly to record, yet many utterances add little new phonetic information. Core-set selection reduces this cost by choosing a small training subset under a fixed audio-duration budget. We represent a corpus as a phonotactic graph that links each utterance to its most phonemically similar ones, and we first test whether this graph has structure. In Bangla and English corpora, its clustering is 199 and 56 times that of a size-matched random graph, and its modularity is more than twice that of a degree-preserving random graph. We then propose Community Representative, a selector that samples across graph communities and spreads its choices within each one, starting from utterances rich in rare phonemes. At every budget and in both languages, it covers more rare phoneme bigrams than random and entropy-based selection, and this lead holds on held-out utterances. TTS models trained on its 20% core-sets have a significantly lower character error rate (CER) than models trained on equal-duration random or entropy-based subsets in both languages. When all models train for the same number of epochs, the Bangla core-set model also outperforms full-corpus training (3.93% vs. 4.47% CER) with 4.5x less training time.