Large language models (LLMs) now dominate NLP and are used in CALL for tasks like grading and error correction, and I will present the ALTA (Cambridge) group’s work comparing LLMs and supervised models, highlighting their strengths, weaknesses, and future directions such as explainability and efficient ‘baby’ LMs.
Researchers: 24 of which 8 PhD students • Research engineers: 12 • Active projects: 9 • Data sets: 1304 • Analyses: 56 • Data points: >34.4G
What we do
Språkbanken Text is a part of Språkbanken, a national e-infrastructure that supports research based on linguistic data.
We develop, refine, and make freely available language data and language technology analyses, with a special focus on the Swedish language throughout history.
We develop freely available digital research platforms, where we aim to support all types of research where language data is central.
We conduct our own research in language technology, including language-based AI, and participate in projects in other disciplines.
More about what we do
We develop, refine, and make freely available language data and language technology analyses, with a special focus on the Swedish language throughout history.
We develop freely available digital research platforms, where we aim to support all types of research where language data is central.
We conduct our own research in language technology, including language-based AI, and participate in projects in other disciplines.
Word of the week
An example of how to use the resources and API:s of Språkbanken Text.
ocean
ocean, Wivallius 1631 = ty. ozean
osv., från grek. ōkeanós, äldst:
världsströmmen kring jordskivan;
sammanställt med sanskr. ā-çáyāna-, den som
omgiver, till ie. roten kei, ligga, vila,
varom under hem , ide (hide).
Svensk etymologisk ordbok, Hellquist, 1922
Fifty years ago, the Logoteket was established at the University of Gothenburg. Today, Språkbanken is a national research infrastructure with extensive national and international collaborations. This makes us one of the world's oldest research and development units in language technology.
Språkbanken was envisioned in an op-ed piece written by Sture Allén for the Swedish daily Dagens Nyheter in September 1970. In 1973, the Computational Linguistics Unit submitted a formal proposal to the Ministry of Education, requesting earmarked funding for what was to become Språkbanken. Two years later, this research infrastructure became a reality, when the Logotheque (as it was called initially) was established with national funding in 1975.