Icelandic NLP resources

This is an list of known tools and resources developed specifically to do linguistic processing in Icelandic. It is intended to give readers a clear overview of the ever-growing arsenal of tools for working with Icelandic natural language data at a glance.

This list is categorized by task to increase clarity. Due to that, some multi-functional tools and toolkits might appear more than once in the list. If you notice a category or resource is missing or have suggestions on how to improve this list, please open a GitHub pull request. For those who do not know how to make pull requests, you can also create an issue with your GitHub account.

Contents

• Notable papers
• Other resource collections
• Corpora
• European Language Grid Services
• Toolkits
• Tokenization and text normalization
• POS tagging
• Syntactic parsing
• Grapheme-to-phoneme
• Stress analysis
• Speech synthesis (TTS)
• Speech recognition (ASR)

Notable papers and reports ↑

• Máltækniáætlun fyrir íslensku 2018-2022 (English version)
  • The project plan for an ongoing language technology programme funded by the Icelandic Ministry of Education.
  • Short paper describing the programme, note that the programme has been postponed by a year compared to the original plan.
• Risamálheild: A Very Large Icelandic Text Corpus
  • Paper describing the Icelandic Gigaword Corpus, a tagged and lemmatized corpus containing over 10^9 tokens.
• A Wide-Coverage Context-Free Grammar for Icelandic and an Accompanying Parsing System
• Please send a pull request with additions to this list. If you create a Github issue with the following details of the paper: title, link/URL to PDF/book, and a short description then we can add it to the website/markdown file.

Other resource collections ↑

• CLARIN-IS
  • The Icelandic branch of the CLARIN-ERIC language resource initiative. Contains information on and downloads for many tools and datasets.
• SÍM homepage
  • Overview page for SÍM (the Icelandic Language Technology Consortium), which contains mirrors and descriptions for all Language Technology Programme projects.
• malfong.is
  • List of language technology resources, maintained by Árnastofnun.
• Comprehensive list of language resources
  • This list of over 100 Icelandic language technology resources was compiled by @bjarnigithub in the summer of 2021.

Corpora ↑

• Talrómur
  • A large public domain TTS corpus designed for research and development. Contains over 160 hours of studio-recorded prompted speech, divided between 8 speakers.
• Samrómur
  • An open and accessible speech recognition dataset with FLAC audio files, corresponding text and metadata.
• Icelandic broadcast speech
  • 193 hours of radio and TV data from the Icelandic National Broadcasting Service (RÚV).
• Spjallromur
  • Icelandic Conversational Speech
• Kennslurómur
  • Icelandic lectures with audio and corresponding text.
• GreynirCorpus
  • A large, parsed treebank of modern Icelandic text

European Language Grid Services ↑

• tokenizer_api
• icenlp_api
• icenlp_api (IceParser - Shallow Parser)
• pos_api
• ner_api
• far_abltagger_api
• icesum_api
• nefnir_api
• greynirseq_api
• greynirseq_api
• binpackage_api
• greynircorrect_api

Toolkits ↑

Greynir

• Python 3 package which is capable of syntactic parsing, lemmatization, POS tagging, noun phrase inflection and more
• The GitHub repo for this project
• Developed by Miðeind ehf.

IceNLP

• Java toolkit which does tokenization, POS tagging, lemmatization, parsing and NER
• Developed by Hrafn Loftsson

LVL-tts-frontend

• TTS frontend designed to work with the Merlin speech synthesis system developed by CSTR
• It contains a pronunciation dictionary, sequitur g2p model, stress analysis component and more. Unfortunately it does not include any documentation.
  • Developed by Anna Björk Nikulásdóttir at LVL

Tokenization and text normalization ↑

• Icelandic tokenizer
• Textahaukur - text normalization toolkit
  • This seems to be in suspended development and claims to not be functional as of yet.
• Regína normalizer
  • Regex-based text normalization in python. Currently in early stages of development.

POS tagging ↑

• POS tagger based on ABLTagger
  • Publication describing ABLTagger
• The standard Icelandic fine-grained POS Tagset

Syntactic parsing ↑

• Neural parsing pipeline for Icelandic
  • The GitHub repo for this project
• Greynir, see above
• IceNLP, see above

Grapheme-to-phoneme ↑

• LSTM encoder-decoder sequence-to-sequence models for Icelandic, reference
• g2p-service is a g2p web service. reference
• Icelandic pronunciation dictionary
• Pronunciation dictionary editor
• Thrax G2P grammar for Icelandic, reference
• LVL-tts-frontend
• G2P - Atli Thor’s g2p python module/pip package, reference
• Module for preparing text data for TTS data collections …, reference
• Althingi ASR g2p, reference

Stress analysis ↑

• LVL-tts-frontend performs stress analysis

Speech synthesis ↑

• Unit selection recipe in Festival for Icelandic
• FastSpeech2 with support for the Talromur corpus

Speech recognition ↑

• Ice-ASR
• Alþingi
  • Just the recipe
• Samromur ASR
  • Contains a vanilla recipe (base), subword modelling, and specialized children and adolescent recipes
• alignment and segmentation
  • Scripts to prepare RÚV TV data for alignment and segmentation to make an ASR dataset
• Tiro Speech Core
• Tal

Our CADIA-LVL works in progress

You can also see our many works in progress at LVL itself if you follow us on our github: https://github.com/cadia-lvl
Facebook page: https://www.facebook.com/languageandvoice/