This overview paper presents the results of the General Machine Translation Task organised as part of the 2024 Conference on Machine Translation (WMT). In the general MT task, participants were asked to build machine translation systems for any of 11 language pairs, to be evaluated on test sets consisting of three to five different domains. In addition to participating systems, we collected translations from 8 different large language models (LLMs) and 4 online translation providers. We evaluate system outputs with professional human annotators using a new protocol called Error Span Annotations (ESA).

This paper presents the submission of the Arni Magnusson Institute’s team to the WMT24 General translation task. We work on the English→Icelandic translation direction. Our system comprises four translation models and a grammar correction model. For training our systems we carefully curate our datasets, aggressively filtering out sentence pairs that may detrimentally affect the quality of our systems output. Some of our data are collected from human translations and some are synthetically generated. A part of the synthetic data is generated using an LLM, and we find that it increases the translation capability of our system significantly.

The submission of the Árni Magnússon Institute’s team to the WMT24 test suite subtask focuses on idiomatic expressions and proper names for the English→Icelandic translation direction. Intuitively and empirically, idioms and proper names are known to be a significant challenge for neural translation models. We create two different test suites. The first evaluates the competency of MT systems in translating common English idiomatic expressions, as well as testing whether systems can distinguish between those expressions and the same phrases when used in a literal context. The second test suite consists of place names that should be translated into their Icelandic exonyms (and correctly inflected) and pairs of Icelandic names that share a surface form between the male and female variants, so that incorrect translations impact meaning as well as readibility. The scores reported are relatively low, especially for idiomatic expressions and place names, and indicate considerable room for improvement.

We present SentAlign, an accurate sentence alignment tool designed to handle very large parallel document pairs. Given user-defined parameters, the alignment algorithm evaluates all possible alignment paths in fairly large documents of thousands of sentences and uses a divide-and-conquer approach to align documents containing tens of thousands of sentences. The scoring function is based on LaBSE bilingual sentence representations. SentAlign outperforms five other sentence alignment tools when evaluated on two different evaluation sets, German-French and English-Icelandic, and on a downstream machine translation task.

This paper describes the AST submission to the WMT23 Shared Task on Parallel Data Curation. We experiment with two approaches for curating data from the provided web-scraped texts. We use sentence alignment to identify document alignments in the data and extract parallel sentence pairs from the aligned documents. All other sentences, not aligned in that step, are paired based on cosine similarity before we apply various different filters. For filtering, we use language detection, fluency classification, word alignments, cosine distance as calculated by multilingual sentence embedding models, and Bicleaner AI. Our best model outperforms the baseline by 1.9 BLEU points on average over the four provided evaluation sets.

When parallel corpora are preprocessed for machine translation (MT) training, a part of the parallel data is commonly discarded and deemed non-parallel due to odd-length ratio, overlapping text in source and target sentences or failing some other form of a semantic equivalency test. For language pairs with limited parallel resources, this can be costly as in such cases modest amounts of acceptable data may be useful to help build MT systems that generate higher quality translations. In this paper, we refine parallel corpora for two language pairs, English–Bengali and English–Icelandic, by extracting sub-sentence fragments from sentence pairs that would otherwise have been discarded, in order to increase recall when compiling training data. We find that by including the fragments, translation quality of NMT systems trained on the data improves significantly when translating from English to Bengali and from English to Icelandic.

We describe the AST submission for the CoCo4MT 2023 shared task. The aim of the task is to identify the best candidates for translation in a source data set with the aim to use the translated parallel data for fine-tuning the mBART-50 model. We experiment with three methods: scoring sentences based on n-gram coverage, using LaBSE to estimate semantic similarity and identify misalignments and mistranslations by comparing machine translated source sentences to corresponding manually translated segments in high-resource languages. We find that we obtain the best results by combining these three methods, using LaBSE and machine translation for filtering, and one of our n-gram scoring approaches for ordering sentences.

We describe work on enhancing the performance of transformer-based encoder-decoder models for OCR post-correction on modern and historical Icelandic texts, where OCRed data are scarce. We trained six models, four from scratch and two fine-tuned versions of Google’s ByT5, on a combination of real data and texts populated with artificially generated errors. Our results show that the models trained from scratch, as opposed to the fine-tuned versions, benefited the most from the addition of artificially generated errors.

We explore different approaches for filtering parallel data for MT training, whether the same filtering approaches suit different datasets, and if separate filters should be applied to a dataset depending on the translation direction. We evaluate the results of different approaches, both manually and on a downstream NMT task. We find that, first, it is beneficial to inspect how well different filtering approaches suit different datasets and, second, that while MT systems trained on data prepared using different filters do not differ substantially in quality, there is indeed a statistically significant difference. Finally, we find that the same training sets do not seem to suit different translation directions.

We present Gamli, an ASR corpus for Icelandic oral histories, the first of its kind for this language, derived from the Ísmús ethnographic collection. Corpora for oral histories differ in various ways from corpora for general ASR, they contain spontaneous speech, multiple speakers per channel, noisy environments, the effects of historic recording equipment, and typically a large proportion of elderly speakers. Gamli contains 146 hours of aligned speech and transcripts, split into a training set and a test set. We describe our approach for creating the transcripts, through both OCR of previous transcripts and post-editing of ASR output. We also describe our approach for aligning, segmenting, and filtering the corpus and finally training a Kaldi ASR system, which achieves 22.4% word error rate (WER) on the Gamli test set, a substantial improvement from 58.4% word error rate from a baseline general ASR system for Icelandic.

We describe the evaluation and development of a rule-based treebank conversion tool, UDConverter, which converts treebanks from the constituency-based PPCHE annotation scheme to the dependency-based Universal Dependencies (UD) scheme. The tool has already been used in the production of three UD treebanks, although no formal evaluation of the tool has been carried out as of yet. By manually correcting new output files from the converter and comparing them to the raw output, we measured the labeled attachment score (LAS) and unlabeled attachment score (UAS) of the converted texts. We obtain an LAS of 82.87 and a UAS of 87.91. In comparison to other tools, UDConverter currently provides the best results in automatic UD treebank creation for Icelandic.

The Icelandic Gigaword Corpus was first published in 2018. Since then new versions have been published annually, containing new texts from additional sources as well as from previous sources. This paper describes the evolution of the corpus in its first four years. All versions are made available under permissive licenses and with each new version the texts are annotated with the latest and most accurate tools. We show how the corpus has grown almost 50% in size from the first version to the fourth and how it was restructured in order to better accommodate different meta-data for different subcorpora. Furthermore, other services have been set up to facilitate usage of the corpus for different use cases. These include a keyword-in-context concordance tool, an n-gram viewer, a word frequency database and pre-trained word embeddings.

Word embedding models have become commonplace in a wide range of NLP applications. In order to train and use the best possible models, accurate evaluation is needed. For extrinsic evaluation of word embedding models, analogy evaluation sets have been shown to be a good quality estimator. We introduce an Icelandic adaptation of a large analogy dataset, BATS, evaluate it on three different word embedding models and show that our evaluation set is apt at measuring the capabilities of such models.

Bilingual lexicons can be generated automatically using a wide variety of approaches. We perform a rigorous manual evaluation of four different methods: word alignments on different types of bilingual data, pivoting, machine translation and cross-lingual word embeddings. We investigate how the different setups perform using publicly available data for the English-Icelandic language pair, doing separate evaluations for each method, dataset and confidence class where it can be calculated. The results are validated by human experts, working with a random sample from all our experiments. By combining the most promising approaches and data sets, using confidence scores calculated from the data and the results of manually evaluating samples from our manual evaluation as indicators, we are able to induce lists of translations with a very high acceptance rate. We show how multiple different combinations generate lists with well over 90% acceptance rate, substantially exceeding the results for each individual approach, while still generating reasonably large candidate lists. All manually evaluated equivalence pairs are published in a new lexicon of over 232,000 pairs under an open license.

Being able to generate accurate word alignments is useful for a variety of tasks. While statistical word aligners can work well, especially when parallel training data are plentiful, multilingual embedding models have recently been shown to give good results in unsupervised scenarios. We evaluate an ensemble method for word alignment on four language pairs and demonstrate that by combining multiple tools, taking advantage of their different approaches, substantial gains can be made. This holds for settings ranging from very low-resource to high-resource. Furthermore, we introduce a new gold alignment test set for Icelandic and a new easy-to-use tool for creating manual word alignments.

The new Icelandic Word Web (IW) is a language technology focused redesign of a lexicosemantic database of semantically related entries. The IW’s entities, relations, metadata and categorization scheme have all been implemented from scratch in two systems, OntoLex and SKOS. After certain adjustments were made to OntoLex and SKOS interoperability, it was also possible to implement specific IW features that, while potentially nonstandard, form an integral part of the Word Web’s lexicosemantic functionality. Also new in this implementation are access to a larger amount of linguistic data, a greater variety of search options, the possibility of automated processing, and the ability to conduct research through SPARQL without possessing a mastery of Icelandic.

Parallel sentences extracted from comparable corpora can be useful to supplement parallel corpora when training machine translation (MT) systems. This is even more prominent in low-resource scenarios, where parallel corpora are scarce. In this paper, we present a system which uses three very different measures to identify and score parallel sentences from comparable corpora. We measure the accuracy of our methods in low-resource settings by comparing the results against manually curated test data for English–Icelandic, and by evaluating an MT system trained on the concatenation of the parallel data extracted by our approach and an existing data set. We show that the system is capable of extracting useful parallel sentences with high accuracy, and that the extracted pairs substantially increase translation quality of an MT system trained on the data, as measured by automatic evaluation metrics.

Automatic term extraction (ATE) from texts is critical for effective terminology work in small speech communities. We present TermPortal, a workbench for terminology work in Iceland, featuring the first ATE system for Icelandic. The tool facilitates standardization in terminology work in Iceland, as it exports data in standard formats in order to streamline gathering and distribution of the material. In the project we focus on the domain of finance in order to do be able to fulfill the needs of an important and large field. We present a comprehensive survey amongst the most prominent organizations in that field, the results of which emphasize the need for a good, up-to-date and accessible termbank and the willingness to use terms in Icelandic. Furthermore we present the ATE tool for Icelandic, which uses a variety of methods and shows great potential with a recall rate of up to 95% and a high C-value, indicating that it competently finds term candidates that are important to the input text.

We describe the acquisition, annotation and encoding of the corpus of the Althingi parliamentary proceedings. The first version of the corpus includes speeches from 1911-2019. It comprises 406 thousand speeches and over 219 million words. The corpus has been automatically part-of-speech tagged and lemmatised. It is annotated with extensive metadata about the speeches, speakers and political parties, including speech topic, whether the speaker is in the government coalition or opposition, age and gender of speaker at the time of delivery, references to sound and video recordings and more. The corpus is encoded in accordance with the Text Encoding Initiative (TEI) Guidelines and conforms to the Parla-CLARIN schema. We plan to update the corpus annually and its major versions will be archived in the CLARIN.IS repository. It is available for download and search using the KORP concordance tool. Furthermore, information on word frequency are accessible in a custom made web application and an n-gram viewer.

LARA (Learning and Reading Assistant) is an open source platform whose purpose is to support easy conversion of plain texts into multimodal online versions suitable for use by language learners. This involves semi-automatically tagging the text, adding other annotations and recording audio. The platform is suitable for creating texts in multiple languages via crowdsourcing techniques that can be used for teaching a language via reading and listening. We present results of initial experiments by various collaborators where we measure the time required to produce substantial LARA resources, up to the length of short novels, in Dutch, English, Farsi, French, German, Icelandic, Irish, Swedish and Turkish. The first results are encouraging. Although there are some startup problems, the conversion task seems manageable for the languages tested so far. The resulting enriched texts are posted online and are freely available in both source and compiled form.

We introduce an array of open and accessible tools to facilitate the use of the Icelandic Gigaword Corpus, in the field of Natural Language Processing as well as for students, linguists, sociologists and others benefitting from using large corpora. A KWIC engine, powered by the Swedish Korp tool is adapted to the specifics of the corpus. An n-gram viewer, highly customizable to suit different needs, allows users to study word usage throughout the period of our text collection. A frequency dictionary provides much sought after information about word frequency statistics, computed for each subcorpus as well as aggregate, disambiguating homographs based on their respective lemmas and morphosyntactic tags. Furthermore, we provide n-grams based on the corpus, and a variety of pre-trained word embeddings models, based on word2vec, GloVe, fastText and ELMo. For three of the model types, multiple word embedding models are available trained with different algorithms and using either lemmatised or unlemmatised texts.

In this paper, we describe a new national language technology programme for Icelandic. The programme, which spans a period of five years, aims at making Icelandic usable in communication and interactions in the digital world, by developing accessible, open-source language resources and software. The research and development work within the programme is carried out by a consortium of universities, institutions, and private companies, with a strong emphasis on cooperation between academia and industries. Five core projects will be the main content of the programme: language resources, speech recognition, speech synthesis, machine translation, and spell and grammar checking. We also describe other national language technology programmes and give an overview over the history of language technology in Iceland.

This contribution describes an ongoing project of speech data collection, using the web application Samrómur which is built upon Common Voice, Mozilla Foundation’s web platform for open-source voice collection. The goal of the project is to build a large-scale speech corpus for Automatic Speech Recognition (ASR) for Icelandic. Upon completion, Samrómur will be the largest open speech corpus for Icelandic collected from the public domain. We discuss the methods used for the crowd-sourcing effort and show the importance of marketing and good media coverage when launching a crowd-sourcing campaign. Preliminary results exceed our expectations, and in one month we collected data that we had estimated would take three months to obtain. Furthermore, our initial dataset of around 45 thousand utterances has good demographic coverage, is gender-balanced and with proper age distribution. We also report on the task of validating the recordings, which we have not promoted, but have had numerous hours invested by volunteers.

Parallel corpora are key to developing good machine translation systems. However, abundant parallel data are hard to come by, especially for languages with a low number of speakers. When rich morphology exacerbates the data sparsity problem, it is imperative to have accurate alignment and filtering methods that can help make the most of what is available by maximising the number of correctly translated segments in a corpus and minimising noise by removing incorrect translations and segments containing extraneous data. This paper sets out a research plan for improving alignment and filtering methods for parallel texts in low-resource settings. We propose an effective unsupervised alignment method to tackle the alignment problem. Moreover, we propose a strategy to supplement state-of-the-art models with automatically extracted information using basic NLP tools to effectively handle rich morphology.

The topic of this paper is a rule-based pipeline for converting constituency treebanks based on the Penn Treebank format to Universal Dependencies (UD). We describe an Icelandic constituency treebank, its annotation scheme and the UD scheme. The conversion is discussed, the methods used to deliver a fully automated UD corpus and complications involved. To show its applicability to corpora in different languages, we extend the pipeline and convert a Faroese constituency treebank to a UD corpus. The result is an open-source conversion tool, published under an Apache 2.0 license, applicable to a Penn-style treebank for conversion to a UD corpus, along with the two new UD corpora.

We present ParIce, a new English-Icelandic parallel corpus. This is the first parallel corpus built for the purposes of language technology development and research for Icelandic, although some Icelandic texts can be found in various other multilingual parallel corpora. We map out which Icelandic texts are available for these purposes, collect aligned data and align other bilingual texts we acquired. We describe the alignment process and how we filter the data to weed out noise and bad alignments. In total we collected 43 million Icelandic words in 4.3 million aligned segment pairs, but after filtering, our corpus includes 38.8 million Icelandic words in 3.5 million segment pairs. We estimate that approximately 5% of the corpus data is noise or faulty alignments while more than 50% of the segments we deleted were faulty. We estimate that our filtering process reduced the number of faulty segments in the corpus by more than 60% while only reducing the number of good alignments by approximately 8%.

The topic of this paper is The Database of Icelandic Morphology (DIM), a multipurpose linguistic resource, created for use in language technology, as a reference for the general public in Iceland, and for use in research on the Icelandic language. DIM contains inflectional paradigms and analysis of word formation, with a vocabulary of approx. 285,000 lemmas. DIM is based on The Database of Modern Icelandic Inflection, which has been in use since 2004.

Previous work on using BiLSTM models for PoS tagging has primarily focused on small tagsets. We evaluate BiLSTM models for tagging Icelandic, a morphologically rich language, using a relatively large tagset. Our baseline BiLSTM model achieves higher accuracy than any other previously published tagger, when not taking advantage of a morphological lexicon. When we extend the model by incorporating such data, we outperform the earlier state-of-the-art results by a significant margin. We also report on work in progress that attempts to address the problem of data sparsity inherent to morphologically detailed, fine-grained tagsets. We experiment with training a separate model on only the lexical category and using the coarse-grained output tag as an input into to the main model. This method further increases the accuracy and reduces the tagging errors by 21.3% compared to previous state-of-the-art results. Finally, we train and test our tagger on a new gold standard for Icelandic.