The KBLab Blog

easytranscriber: Speech recognition with precise timestamps

easytranscriber is an automatic speech recognition library for transcription with precise word-level timestamps. It is 35% to 102% faster compared to WhisperX, by leveraging…

Faton Rekathati

Feb 26, 2026

CLIP-Topic: Identifying Themes in Large Image Collections

Heritage institutions hold vast collections of digital images that remain difficult to search or interpret. CLIP-Topic shows how multimodal AI can be used to surface…

Chris Haffenden, Justyna Sikora

Dec 1, 2025

From Parchment to Pixels: Testing HTR for Medieval Latin Manuscripts at KBLab

How well do current HTR models handle the complexity of medieval Latin handwriting? At KBLab, we’ve been testing and comparing the state of the art in openly available…

Justyna Sikora, Chris Haffenden, Robin Böckerman

Jun 11, 2025

Welcome KB-Whisper, a new fine-tuned Swedish Whisper model!

KBLab proudly presents KB-Whisper, a speech to text model fine-tuned using 50,000 hours of transcribed speech. This work reports an overall improvement across model sizes…

Leonora Vesterbacka, Faton Rekathati, Robin Kurtz, Justyna Sikora, Agnes Toftgård

Mar 7, 2025

Preserving the history of cultural heritage conservation

While KBLab uses modern heritage data to produce cutting edge AI models, we also work with developing the library’s research infrastructure for older heritage material and…

Chris Haffenden, Emil Stenback

Jan 31, 2025

Analysis of financial data at the Financial Supervisory Authority

A guest post from the Swedish Financial Supervisory Authority (Finansinspektionen, FI), where FI describe how they use machine learning to analyze financial data and perform…

Jimmy Hollén

Jan 30, 2024

Unearthing forgotten images with the help of AI

How can new AI methods be used to improve the searchability and accessibility of visual heritage collections? We’ve taken advantage of the possibilities opened up by…

Chris Haffenden, Faton Rekathati, Emma Rende

Oct 20, 2023

Words unboxed: discovering new words with Kubord

KBLab together with Språkbanken Text have developed 75 freely available datasets to support research in, but not limited to, lexicography. The datasets, marked as Kubord 2…

Markus Forsberg, Justyna Sikora, Emma Sköldberg

Aug 29, 2023

For how long is a person recognisable by their voice?

Searching a database of speakers by their voice presents a unique challenge, as speakers’ voices change as they age. We can represent a speaker’s voice computationally with…

Maya Nachesa

Jul 4, 2023

A robust, multi-label sentiment classifier for Swedish

KBLab presents a robust, multi-label sentiment classifier trained on Swedish texts. The model is robust in the sense that it is trained on multiple datasets of different…

Hillevi Hägglöf

Jun 16, 2023

Swedish speech synthesis

KBLab releases a neural network based text-to-speech model for Swedish. The model was trained on an open Swedish speech synthesis dataset from NST. We make our latest…

Faton Rekathati

May 24, 2023

Scientific discourse with BERTopic

We describe a typical topic modeling use case where BERTopic is applied to scientific abstracts in the research field of education. We discuss the limitations of BERTopic…

Hillevi Hägglöf, Justyna Sikora

Mar 30, 2023

RixVox: A Swedish Speech Corpus with 5500 Hours of Speech from Parliamentary Debates

KBLab releases RixVox, a speech dataset comprised of 5500 hours of speech from parliamentary debates. The speeches have been aligned with transcripts from written protocols…

Faton Rekathati

Mar 9, 2023

Finding Speeches in the Riksdag’s Debates

The Riksdag is the Parliament of Sweden. It has made available twenty years of parliamentary debates through its website and open data platform. Each speech is accompanied…

Faton Rekathati

Feb 23, 2023

Swedish zero-shot classification model

KBlab has released a BERT model fine-tuned on NLI tasks, which can be used for zero-shot text classification.

Justyna Sikora

Feb 13, 2023

Swedish Sentence Transformer 2.0

KBLab’s Swedish sentence transformer has been updated to a newer version. The new version features an increased maximum sequence length of 384 tokens, allowing users to…

Faton Rekathati

Jan 16, 2023

BERTopic for Swedish: Topic modeling made easier via KB-BERT

Topic modeling is an exciting option for exploring and finding patterns in large volumes of text data. While this previously required considerable programming skills, a…

Elena Fano, Chris Haffenden

Jun 14, 2022

Evaluating Swedish Language Models

We present OverLim, a new benchmark for evaluating large language models for Swedish, Danish, and Norwegian, created by translating a subset of the GLUE and SuperGLUE…

Robin Kurtz

Mar 16, 2022

Swedish Bootleg model

We at KBLab have trained a Swedish version of an entity disambiguation model called Bootleg, developed by the Hazy Research Lab at Stanford. The model is trained on Swedish…

Elena Fano

Mar 3, 2022

SUCX 3.0 - NER

We present a remix of the venerable SUC 3.0 dataset for Swedish Named Entity Recognition (NER), and explore the effect of Hyper Parameter Optimization (HPO) for this task…

Robin Kurtz, Joey Öhman

Feb 7, 2022

KBLab publishes an article about AI in the library in C&RL

What does AI mean for libraries and how could libraries pave the way for ethical AI? KBLab reflects on these questions in an article in the Open Access journal College &…

Elena Fano, Chris Haffenden

Jan 24, 2022

KBLab’s director Love Börjeson nominated as AI Swede of the year 2021

We are very pleased to announce that our director Love Börjeson has been nominated for the prestigious prize Årets AI Svensk 2021 (AI Swede of the year 2021) which is…

Elena Fano

Nov 15, 2021

Introducing a Swedish Sentence Transformer

While language models such as BERT are effective at many tasks, they have limited use when it comes to information retrieval and large scale similarity comparisons. In this…

Faton Rekathati

Aug 23, 2021

A Swedish-Norwegian Federated Language Model

We trained a bilingual Swedish-Norwegian ELECTRA language model in a federated setup, showcasing LM training when various…

Robin Kurtz

Jun 15, 2021

Topic models for Statens Offentliga Utredningar

We built some topic models on the National Library’s SOU collection, as an example of what can be done with the Library’s materials. The curated dataset can be a useful…

Elena Fano

May 4, 2021

A multimodal approach to advertisement classification in digitized newspapers

The process of digitizing historical newspapers at the National Library of Sweden involves scanning physical copies of newspapers and storing them as images. In order to…

Faton Rekathati

Mar 28, 2021