mlsanityData to 7 Oct 2026

Reusing Latent Speech Representations for Query-Conditioned Topic Localization in Transcripts

arXiv 2609.21844 ↗First posted 18 Sep 2026cs.CL eess.AScited by 0

Not on the map. Why?

Abstract (arXiv)

Long transcripts are costly inputs for downstream NLP systems and often contain irrelevant context. We study query-conditioned topic localization: predicting the sentence span in a transcript that best addresses a topic-title query. To improve span localization, we reuse ASR encoder states as sentence-level representations and fuse them with textual embeddings. This lets lightweight span locators exploit speech information without running a separate audio encoder. Experiments on two public datasets show consistent gains over text-only baselines, especially under strict boundary-matching criteria. Cross-dataset experiments further indicate that the benefits are strongest for structured or semi-structured speech, while gains on spontaneous speech are limited and mixed.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 30 of its 47 references that are in the map. First posted on arXiv 18 Sep 2026.

Builds on

30 of 47 references.

References · 30

  1. 2020wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representationscited by 3,853
  2. 2023Lost in the Middle: How Language Models Use Long Contextscited by 3,383
  3. 2025Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Modelscited by 1,175
  4. 2015Pointer Networkscited by 882
  5. 2021SUPERB: Speech processing Universal PERformance Benchmarkcited by 652
  6. 2020Making Monolingual Sentence Embeddings Multilingual using Knowledge Distillationcited by 591
  7. 2021QMSum: A New Benchmark for Query-based Multi-domain Meeting Summarizationcited by 363
  8. 2024Multilingual E5 Text Embeddings: A Technical Reportcited by 357
  9. 2023WhisperX: Time-Accurate Speech Transcription of Long-Form Audiocited by 320
  10. 2021Layer-wise Analysis of a Self-supervised Speech Representation Modelcited by 293
  11. 2024Long-context LLMs Struggle with Long In-context Learningcited by 285
  12. 2017Gated Multimodal Units for Information Fusioncited by 244
18 more
  1. 2020Span-based Localizing Network for Natural Language Video Localizationcited by 234
  2. 2025NoLiMa: Long-Context Evaluation Beyond Literal Matchingcited by 120
  3. 2018Text Segmentation as a Supervised Learning Taskcited by 76
  4. 2023Whisper-AT: Noise-Robust Automatic Speech Recognizers are Also Strong General Audio Event Taggerscited by 72
  5. 2021Natural Language Video Localization: A Revisit in Span-based Question Answering Frameworkcited by 55
  6. 2020Text Segmentation by Cross Segment Attentioncited by 54
  7. 2025Canary-1B-v2 & Parakeet-TDT-0.6B-v3: Efficient and High-Performance Models for Multilingual ASR and ASTcited by 51
  8. 2016Machine Comprehension Using Match-LSTM and Answer Pointercited by 47
  9. 2023VidChapters-7M: Video Chapters at Scalecited by 39
  10. 2021Membership Inference Attacks Against Self-supervised Speech Modelscited by 16
  11. 2024From Text Segmentation to Smart Chaptering: A Novel Benchmark for Structuring Video Transcriptionscited by 16
  12. 2023Investigating Pre-trained Audio Encoders in the Low-Resource Conditioncited by 11
  13. 2020Improving automated segmentation of radio shows with audio embeddingscited by 4
  14. 2026Towards Multi-Level Transcript Segmentation: LoRA Fine-Tuning for Table-of-Contents Generationcited by 4
  15. 2025Eta-WavLM: Efficient Speaker Identity Removal in Self-Supervised Speech Representations Using a Simple Linear Equationcited by 3
  16. 2026Beyond Transcripts: A Renewed Perspective on Audio Chapteringcited by 3
  17. 2024Advancing Topic Segmentation of Broadcasted Speech with Multilingual Semantic Embeddingscited by 1
  18. 2026Reading Between the Waves: Robust Topic Segmentation Using Inter-Sentence Audio Featurescited by 1

Built on by

Not cited yet.