Abstract
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.
Keywords
Subject
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Freisinger, S., Seeberger, P., Ranzenberger, T., Bocklet, T., & Riedhammer, K. (2026). Reusing Latent Speech Representations for Query-Conditioned Topic Localization in Transcripts. https://omanscience.com/en/articles/reusing-latent-speech-representations-for-query-conditioned-topic-localization-in-transcripts
MLA 9
Freisinger, Steffen, et al. "Reusing Latent Speech Representations for Query-Conditioned Topic Localization in Transcripts." https://omanscience.com/en/articles/reusing-latent-speech-representations-for-query-conditioned-topic-localization-in-transcripts.
Chicago (author–date)
Freisinger, Steffen, Philipp Seeberger, Thomas Ranzenberger, Tobias Bocklet, and Korbinian Riedhammer. 2026. "Reusing Latent Speech Representations for Query-Conditioned Topic Localization in Transcripts." https://omanscience.com/en/articles/reusing-latent-speech-representations-for-query-conditioned-topic-localization-in-transcripts.
Harvard
Freisinger, S., Seeberger, P., Ranzenberger, T., Bocklet, T. and Riedhammer, K. (2026) 'Reusing Latent Speech Representations for Query-Conditioned Topic Localization in Transcripts', Available at: https://omanscience.com/en/articles/reusing-latent-speech-representations-for-query-conditioned-topic-localization-in-transcripts.
Vancouver
Freisinger S, Seeberger P, Ranzenberger T, Bocklet T, Riedhammer K. Reusing Latent Speech Representations for Query-Conditioned Topic Localization in Transcripts. https://omanscience.com/en/articles/reusing-latent-speech-representations-for-query-conditioned-topic-localization-in-transcripts
IEEE
S. Freisinger, P. Seeberger, T. Ranzenberger, T. Bocklet, and K. Riedhammer, "Reusing Latent Speech Representations for Query-Conditioned Topic Localization in Transcripts," https://omanscience.com/en/articles/reusing-latent-speech-representations-for-query-conditioned-topic-localization-in-transcripts.