This paper describes Slav-NER: the 3rd Multilingual Named Entity Challenge in Slavic languages. The tasks involve recognizing mentions of named entities in Web documents, normalization of the names, and cross-lingual linking. The Challenge covers six languages and five entity types, and is organized as part of the 8th Balto-Slavic Natural Language Processing Workshop, co-located with the EACL 2021 Conference. Ten teams participated in the competition. Performance for the named entity recognition task reached 90% F-measure, much higher than reported in the first edition of the Challenge. Seven teams covered all six languages, and five teams participated in the cross-lingual entity linking task. Detailed valuation information is available on the shared task web page.

The paper presents an unsupervised method for quickly extending a Ukrainian lexicon by generating paradigms and morphological feature structures for new Named Entities and neologisms, which are not covered by existing static morphological resources. This approach addresses a practical problem of modelling paradigms for entities created by the dynamic processes in the lexicon: this problem is especially serious for highly-inflected languages in domains with specialised or quickly changing lexicon. The method uses an unannotated Ukrainian corpus and a small fixed set of inflection tables, which can be found in traditional grammar textbooks. The advantage of the proposed approach is that updating the morphological lexicon does not require training or linguistic annotation, allowing fast knowledge-light extension of an existing static lexicon to improve morphological coverage on a specific corpus. The method is implemented in an open-source package on a GitHub repository. It can be applied to other low-resourced inflectional languages which have internet corpora and linguistic descriptions of their inflection system, following the example of inflection tables for Ukrainian. Evaluation results shows consistent improvements in coverage for Ukrainian corpora of different corpus types.

The paper presents a methodology for the development and task-based evaluation of phonological models, which improve the accuracy of cognate terminology identification, but may potentially be used for other applications, such as transliteration or improving character-based NMT. Terminology translation remains a bottleneck for MT, especially for under-resourced languages and domains, and automated identification of cognate terms addresses this problem. The proposed phonological models explicitly represent distinctive phonological features for each character, such as acoustic types (e.g., vowel/ consonant, voiced/ unvoiced/ sonant), place and manner of articulation (closed/open, front/back vowel; plosive, fricative, or labial, dental, glottal consonant). The advantage of such representations is that they explicate information about characters’ internal structure rather than treat them as elementary atomic units of comparison, placing graphemes into a feature space that provides additional information about their articulatory (pronunciation-based) or acoustic (soundbased) distances and similarity. The article presents experimental results of using the proposed phonological models for extracting cognate terminology with the phonologically aware Levenshtein edit distance, which for Top-1 cognate ranking metric outperforms the baseline character-based Levenshtein by 16.5%. Project resources are released on: https://github.com/bogdanbabych/cognates-phonology

In this paper we introduce MoBiL, a hybrid Monolingual, Bilingual and Language modelling feature set and feature selection and evaluation framework. The set includes translation quality indicators that can be utilized to automatically predict the quality of human translations in terms of content adequacy and language fluency. We compare MoBiL with the QuEst baseline set by using them in classifiers trained with support vector machine and relevance vector machine learning algorithms on the same data set. We also report an experiment on feature selection to opt for fewer but more informative features from MoBiL. Our experiments show that classifiers trained on our feature set perform consistently better in predicting both adequacy and fluency than the classifiers trained on the baseline feature set. MoBiL also performs well when used with both support vector machine and relevance vector machine algorithms.

In this paper we present a metric that measures comparability of documents across different languages. The metric is developed within the FP7 ICT ACCURAT project, as a tool for aligning comparable corpora on the document level; further these aligned comparable documents are used for phrase alignment and extraction of translation equivalents, with the aim to extend phrase tables of statistical MT systems without the need to use parallel texts. The metric uses several features, such as lexical information, document structure, keywords and named entities, which are combined in an ensemble manner. We present the results by measuring the reliability and effectiveness of the metric, and demonstrate its application and the impact for the task of parallel phrase extraction from comparable corpora.

The extraction of dictionaries from parallel text corpora is an established technique. However, as parallel corpora are a scarce resource, in recent years the extraction of dictionaries using comparable corpora has obtained increasing attention. In order to find a mapping between languages, almost all approaches suggested in the literature rely on a seed lexicon. The work described here achieves competitive results without requiring such a seed lexicon. Instead it presupposes mappings between comparable documents in different languages. For some common types of textual resources (e.g. encyclopedias or newspaper texts) such mappings are either readily available or can be established relatively easily. The current work is based on Wikipedias where the mappings between languages are determined by the authors of the articles. We describe a neural-network inspired algorithm which first characterizes each Wikipedia article by a number of keywords, and then considers the identification of word translations as a variant of word alignment in a noisy environment. We present results and evaluations for eight language pairs involving Germanic, Romanic, and Slavic languages as well as Chinese.

Lack of sufficient parallel data for many languages and domains is currently one of the major obstacles to further advancement of automated translation. The ACCURAT project is addressing this issue by researching methods how to improve machine translation systems by using comparable corpora. In this paper we present tools and techniques developed in the ACCURAT project that allow additional data needed for statistical machine translation to be extracted from comparable corpora. We present methods and tools for acquisition of comparable corpora from the Web and other sources, for evaluation of the comparability of collected corpora, for multi-level alignment of comparable corpora and for extraction of lexical and terminological data for machine translation. Finally, we present initial evaluation results on the utility of collected corpora in domain-adapted machine translation and real-life applications.

The exploitation of large corpora to create and populate shared translation resources has been hampered in two areas: first, practical problems (“locked-in” data, ineffective exchange formats, client reservations); and second, ethical and legal problems. Recent developments, notably online collaborative translation environments (Desillets, 2007) and greater industry openness, might have been expected to highlight such issues. Yet the growing use of shared data is being addressed only gingerly. Good reasons lie behind the failure to broach the ethics of shared resources. The issues are challenging: confidentiality, ownership, copyright, authorial rights, attribution, the law, protectionism, costs, fairness, motivation, trust, quality, reliability. However, we argue that, though complex, these issues should not be swept under the carpet. The huge demand for translation cannot be met without intelligent sharing of resources (Kelly, 2009). Relevant ethical considerations have already been identified in translation and related domains, in such texts as Codes of Ethics, international conventions and declarations, and Codes of Professional Conduct; these can be useful here. We outline two case studies from current industry initiatives, highlighting their ethical implications. We identify questions which users and developers should be asking and relate these to existing debates and codes as a practical framework for their consideration.

We report on an on-going research project aimed at increasing the range of translation equivalents which can be automatically discovered by MT systems. The methodology is based on semi-supervised learning of indirect translation strategies from large comparable corpora and applying them in run-time to generate novel, previously unseen translation equivalents. This approach is different from methods based on parallel resources, which currently can reuse only individual translation equivalents. Instead it models translation strategies which generalise individual equivalents and can successfully generate an open class of new translation solutions. The task of the project is integration of the developed technology into open-source MT systems.

We report the results of our experiment on assessing the ability of automated MT evaluation metrics to remain sensitive to variations in MT quality as the average quality of the compared systems goes up. We compare two groups of metrics: those, which measure the proximity of MT output to some reference translation, and those which evaluate the performance of some automated process on degraded MT output. The experiment shows that proximity-based metrics (such as BLEU) loose sensitivity as the scores go up, but performance-based metrics (e.g., Named Entity recognition from MT output) remain sensitive across the scale. We suggest a model for explaining this result, which attributes stable sensitivity of performance-based metrics to measuring cumulative functional effect of different language levels, while proximity-based metrics measure structural matches on a lexical level and therefore miss higher-level errors that are more typical for better MT systems. Development of new automated metrics should take into account possible decline in sensitivity on higher-quality MT, which should be tested as part of meta-evaluation of the metrics.

In this paper we present a tool for finding appropriate translation equivalents for words from the general lexicon using comparable corpora. For a phrase in the source language the tool suggests arange of possible expressions used in similar contexts in target language corpora. In the paper we discuss the method and present results of human evaluation of the performance of the tool.

The use of n-gram metrics to evaluate the output of MT systems is widespread. Typically, they are used in system development, where an increase in the score is taken to represent an improvement in the output of the system. However, purchasers of MT systems or services are more concerned to know how well a score predicts the acceptability of the output to a reader-user. Moreover, they usually want to know if these predictions will hold across a range of target languages and text types. We describe an experiment involving human and automated evaluations of four MT systems across two text types and 23 language directions. It establishes that the correlation between human and automated scores is high, but that the predictive power of these scores depends crucially on target language and text type.