Pattern recognition (psychology)

Speech segmentation

Feature (linguistics)

Mathematics

Text segmentation

Philosophy

Artificial intelligence

Geometry

Phone

Word error rate

Computer science

Linguistics

Cluster analysis

Speech recognition

Artificial neural network

Feature vector

Segmentation

Word (group theory)

We investigate segmenting and clustering speech into low-bitrate phone-like sequences without supervision.We specifically constrain pretrained self-supervised vector-quantized (VQ) neural networks so that blocks of contiguous feature vectors are assigned to the same code, thereby giving a variable-rate segmentation of the speech into discrete units.Two segmentation methods are considered.In the first, features are greedily merged until a prespecified number of segments are reached.The second uses dynamic programming to optimize a squared error with a penalty term to encourage fewer but longer segments.We show that these VQ segmentation methods can be used without alteration across a wide range of tasks: unsupervised phone segmentation, ABX phone discrimination, same-different word discrimination, and as inputs to a symbolic word segmentation algorithm.The penalized dynamic programming method generally performs best.While performance on individual tasks is only comparable to the state-of-the-art in some cases, in all tasks a reasonable competing approach is outperformed at a substantially lower bitrate.

Towards Unsupervised Phone and Word Segmentation Using Self-Supervised Vector-Quantized Neural Networks