Telefónica Innovación Digital 公布 Speech-COSER Challenge 2026 ASR 赛道方案
TID System for Speech-COSER Challenge 2026: ASR Track
Telefónica 这套方案专门啃老年人口语识别这块硬骨头，用 CTC 对齐打分来安排 LoRA 微调顺序，再剪枝防过拟合，7 个系统融合后归一化 WER 做到 22.17%。做低资源语音或适老产品的工程师可以直接抄作业。
Telefónica Innovación Digital 公开其参加 Speech-COSER Challenge 2026 ASR 赛道的系统方案，比赛数据为西班牙农村地区老年说话者的自发性语音。方案核心是 CTC 伪强制对齐，用音频-文本匹配质量打分来筛选数据，指导两阶段 LoRA 课程学习，再对 LoRA 矩阵做事后幅度剪枝以提升泛化。最终由 7 个系统通过 ROVER 融合，取得 36.25% 非归一化 WER 和 22.17% 归一化 WER。
技术路径上，伪强制对齐解决了老年自发言语中标注噪声大的问题：CTC 对齐得分低的样本在 LoRA 微调课程中被降权或延后。两阶段课程先学通用适配、再精调难样本，剪枝则压缩 LoRA 参数防止过拟合小规模方言口音数据。
这类面向低资源、高龄说话人群体的 ASR 方案对适老化语音产品、方言数据微调团队有直接参考价值，ROVER 多系统融合也说明单一 LoRA 微调在该赛道尚未逼近上限。
TID System for Speech-COSER Challenge 2026: ASR Track
We describe the system submitted by Telefónica Innovación Digital to the Speech-COSER Challenge 2026, which features spontaneous speech from elderly speakers recorded in rural Spain. For ASR, we perform CTC-based pseudo-forced alignment scores audio-text match quality to guide a two-stage LoRA curriculum, followed by post-hoc magnitude pruning of the LoRA matrices to enhance generalization. An ensemble of seven systems combined via ROVER achieves 36.25\% unnormalized and 22.17\% normalized WER.
This report is categorized under AITOP's "Papers" desk, focusing on Europe's AI governance and technology ecosystem. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include ASR, LoRA, CTC, 语音识别.
Source attribution: curated from arXiv cs.AI (@Fernando López, Jordi Luque).
