arXiv Computation and Language

Recovering the Zipfian Distribution in Unsupervised Term Discovery

The paper investigates unsupervised term discovery in speech, comparing centre-based clustering methods like K‑means with graph‑based clustering using the Leiden algorithm. It finds that graph clustering produces lexicons whose type frequencies follow a Zipfian distribution, outperforming K‑means, GMM, and BIRCH across word‑ and syllable‑level discovery in three languages. Agglomerative clustering with average linkage also performs well but is less efficient and offers less control over the distribution.

arXiv Computation and Language
Sep 22

Revisiting Lexicon Evaluation in Unsupervised Word Discovery

The paper critiques the normalized edit distance metric used for evaluating lexicons derived from unsupervised word discovery, noting its bias toward large clusters and its failure to account for the distribution of true classes across clusters. It proposes two new metrics—one that weights cluster size when measuring within‑cluster consistency and another that evaluates how true words are spread across clusters—drawing on clustering theory. Experiments on synthetic and real‑world lexicons show that these combined metrics better correlate with ground‑truth distributions and are more robust to evaluation biases.

By Simon Malan, Danel Slabbert, Herman Kamper
arXiv AI
Jun 17

findsylls: A Language-Agnostic Toolkit for Syllable-Level Speech Tokenization and Embedding

arXiv:2603. 26292v2 Announce Type: replace-cross Abstract: Syllable-level units offer compact and linguistically meaningful representations for spoken language modeling and unsupervised word discovery, but research on syllabification remains fragmented across disparate implementations, datasets, and evaluation protocols.

By H\'ector Javier V\'azquez Mart\'inez
arXiv AI
Aug 20

SuTRA : Structurally-Unified Tokenization with Root Awareness

SuTRA (Structurally-Unified Tokenization with Root Awareness) is a morphology-aware tokenization algorithm designed to address the problem of Morphological Shattering in morphologically rich Indic languages. It preserves the indivisibility of aksharas—complex orthographic syllables—by penalizing merges that cross morphological boundaries, thereby reducing over-fragmentation of words. The authors also release a new morphological segmentation dataset for Hindi, Marathi, and Gujarati, and demonstrate that SuTRA improves morphological alignment by up to 14.7% and semantic recoverability by 34% over BPE, leading to an average machine translation gain of +8.08 chrF2.

By Vaibhav Rathore, Siddhant Gole, Dadhichi Telwadkar, Rooshil Bhatia, Maulik Ruparel, Siddharth Surekha, Neha Bhargava
arXiv Computation and Language
5d ago

DiscoPhon: Benchmarking the Unsupervised Discovery of Phoneme Inventories With Discrete Speech Units

DiscoPhon is a multilingual benchmark designed to evaluate unsupervised phoneme discovery from discrete speech units. It includes 6 development and 6 test languages that cover a wide range of phonemic contrasts, and requires systems to generate discrete units mapped to a predefined phoneme inventory using only 10 hours of speech from an unseen language. The benchmark assesses unit quality, recognition, and segmentation, and provides four pretrained multilingual HuBERT and SpidR baselines that demonstrate current models can produce units that correlate well with phonemes, though performance varies across languages.

By Maxime Poli, Manel Khentout, Angelo Ortiz Tandazo, Ewan Dunbar, Emmanuel Chemla, Emmanuel Dupoux
arXiv Computation and Language
Sep 18

Dictionary-Constrained Grapheme-to-Phoneme for Unsegmented Languages from LLM-Annotated Data

The paper introduces a context‑aware neural grapheme‑to‑phoneme (G2P) system for unsegmented languages like Japanese, using a discriminative conditional random field over a word lattice built from dictionaries. It addresses data scarcity by generating over two million sentences with large language models. Experiments show the method surpasses traditional morphological analyzers and neural sequence models, achieving 99.62% target word reading accuracy and very low phoneme error rates on the Joyo‑Kanji‑Yomi benchmark.

By Rui Hu, Zhenpeng Zhan, Xiaolong Lin