arxiv:2610.01450

Code-Switching Spoken Language Identification as Multi-Label Set Prediction

Published on Oct 1

Authors:

Abstract

Code-switched speech evades monolingual language identification filters, prompting a multi-label set-generation approach for code-switch-aware LID that reveals obstacles including cardinality estimation, threshold instability, data bias, and synthetic-to-real gaps.

Generated by thinkingmachines/Inkling-Small

Code-switched (CS) speech leaks through the monolingual language identification (LID) filters used to curate massive speech corpora, calling for CS-aware LID (CS-LID). We formulate utterance-level CS-LID as multi-label language-set prediction and propose a set generator that directly outputs the languages in an utterance, comparing it against atomic-pair and score-based classification baselines. Oracle Top-k is the strongest baseline, but thresholding fails because no single threshold separates CS from monolingual speech. Our set generator predicts the correct language count on unseen pairs without assuming the number of languages, but underperforms oracle Top-k in exact set accuracy. Our analysis identifies the key obstacles to robust CS-LID: oracle cardinality, threshold instability, language bias in CS training data, and the synthetic-to-real gap.