Search

This deliverable presents a method to provide the researchers with a consistent software environment across multiple computing facilities. The consistency is achieved by constructing a set of build scripts that will install the software tools with the same configurations and versions with the same set of dependencies. EasyBuild was selected as the framework to install and configure the software from source code, optimised for the underlying system and expose them as software-modules with uniform naming across systems. This way the researchers can bring their analysis pipelines and run on any of the systems just by loading the same set of modules., This deliverable is approved by the EC on 2nd March 2023. The EOSC-Nordic project has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 857652.

We describe the second IWPT task on end-to-end parsing from raw text to Enhanced Universal Dependencies. We provide details about the evaluation metrics and the datasets used for training and evaluation. We compare the approaches taken by participating teams and discuss the results of the shared task, also in comparison with the first edition of this task.

We present the ongoing NorLM initiative to support the creation and use of very large contextualised language models for Norwegian (and in principle other Nordic languages), including a ready-to-use software environment, as well as an experience report for data preparation and training. This paper introduces the first large-scale monolingual language models for Norwegian, based on both the ELMo and BERT frameworks. In addition to detailing the training process, we present contrastive benchmark results on a suite of NLP tasks for Norwegian. For additional background and access to the data, models, and software, please see http://norlm.nlpl.eu, Comment: Accepted to NoDaLiDa'2021

The 2020 Shared Task at the Conference for Computational Language Learning (CoNLL) was devoted to Meaning Representation Parsing (MRP) across frameworks and languages. Extending a similar setup from the previous year, five distinct approaches to the representation of sentence meaning in the form of directed graphs were represented in the English training and evaluation data for the task, packaged in a uniform graph abstraction and serialization; for four of these representation frameworks, additional training and evaluation data was provided for one additional language per framework. The task received submissions from eight teams, of which two do not participate in the official ranking because they arrived after the closing deadline or made use of additional training data. All technical information regarding the task, including system submissions, official results, and links to supporting resources and software are available from the task web site at: http://mrp.nlpl.eu

The 2020 Shared Task at the Conference for Computational Language Learning (CoNLL) was devoted to Meaning Representation Parsing (MRP) across frameworks and languages. Extending a similar setup from the previous year, five distinct approaches to the representation of sentence meaning in the form of directed graphs were represented in the English training and evaluation data for the task, packaged in a uniform graph abstraction and serialization; for four of these representation frameworks, additional training and evaluation data was provided for one additional language per framework. The task received submissions from eight teams, of which two do not participate in the official ranking because they arrived after the closing deadline or made use of additional training data. All technical information regarding the task, including system submissions, official results, and links to supporting resources and software are available from the task web site at: http://mrp.nlpl.eu

Discourse Representation Theory (DRT) is a formal account for representing the meaning of natural language discourse. Meaning in DRT is modeled via a Discourse Representation Structure (DRS), a meaning representation with a model-theoretic interpretation, which is usually depicted as nested boxes. In contrast, a directed labeled graph is a common data structure used to encode semantics of natural language texts. The paper describes the procedure of dressing up DRSs as directed labeled graphs to include DRT as a new framework in the 2020 shared task on Cross-Framework and Cross-Lingual Meaning Representation Parsing. Since one of the goals of the shared task is to encourage unified models for several semantic graph frameworks, the conversion procedure was biased towards making the DRT graph framework somewhat similar to other graph-based meaning representation frameworks., Comment: 10 pages, 4 figures, 4 tables, CoNLL 2020 Shared Task

We describe the Saarland University submission to the shared task on Cross-Framework Meaning Representation Parsing (MRP) at the 2019 Conference on Computational Natural Language Learning (CoNLL).

We describe the Saarland University submission to the shared task on Cross-Framework Meaning Representation Parsing (MRP) at the 2019 Conference on Computational Natural Language Learning (CoNLL).

In this paper, we empirically evaluate the utility of transfer and multi-task learning on a challenging semantic classification task: semantic interpretation of noun--noun compounds. Through a comprehensive series of experiments and in-depth error analysis, we show that transfer learning via parameter initialization and multi-task learning via parameter sharing can help a neural classification model generalize over a highly skewed distribution of relations. Further, we demonstrate how dual annotation with two distinct sets of relations over the same set of compounds can be exploited to improve the overall accuracy of a neural classifier and its F1 scores on the less frequent, but more difficult relations., EMNLP 2018: Conference on Empirical Methods in Natural Language Processing (EMNLP)

This paper describes an emerging shared repository of large-text resources for creating word vectors, including pre-processed corpora and pre-trained vectors for a range of frameworks and configurations. This will facilitate reuse, rapid experimentation, and replicability of results.

In this contribution, we deal with graph parsing, i.e., mapping input strings to graph-structured output representations, using tree approximations. We experiment with the data from the SemEval 2014 Semantic Dependency Parsing (SDP) task. We define various tree approximation schemes for graphs, and make twofold use of them. First, we statically analyze the semantic dependency graphs, seeking to unscover which linguistic phenomena in particular require the additional annotation expressivity provided by moving from trees to graphs. We focus on undirected base cycles in the SDP graphs, and discover strong connections to grammatical control and coordination. Second, we make use of the approximations in a statistical parsing scenario. In it, we convert the training set graphs to dependency trees, and use the resulting treebanks to build standard dependency tree parsers. We perform lossy graph reconstructions on parser outputs, and evaluate our models as dependency graph parsers. Our system outperforms the baselines by a large margin, and evaluates as the best non-voting tree approximation–based parser on the SemEval 2014 data, scoring at just over 81% in labeled F1.

This demonstration presents a first operable pilot of the Language Analysis Portal (LAP), an ongoing project within the Norwegian CLARINO initiative that aims at providing easy access to Language Technology (LT) tools running on a powerful High-Performance Computing (HPC) cluster. The system is built on top of the Galaxy framework, giving users an on-line platform where they can design experiments using an array of processors. These processors can be combined into complex workflows using a visual editor. The current implementation functions as a testbed for further development, hosting a limited collection of tools addressing common use-cases in the LT-realm; the long-term goal for LAP is to reach beyond the field and be an enabling platform for LT-powered research in the humanities and social sciences. Emanuele Lapponi, Erik Velldal, Nikolay A. Vazov, Stephan Oepen (2013). HPC-ready Language Analysis for Human Beings, Proceedings of the 19th Nordic Conference of Computational Linguistics (NODALIDA 2013), May 22–24, 2013, Oslo University, Norway. NEALT Proceedings Series 16 http://www.ep.liu.se/ecp_article/index.en.aspx?issue=085;article=042

This paper documents ongoing work within the Norwegian CLARINO project on building a Language Analysis Portal (LAP). The portal will provide an intuitive and easily accessible web interface to a centralized repository of a wide range of language technology tools, all installed on a high-performance computing cluster. Users will be able to compose and run workflows using an easy-to-use graphical interface, with multiple tools and resources chained together in potentially complex pipelines. Although the project aims to reach out to a diverse set of user groups, it particularly will facilitate use of language analysis in the social sciences, humanities, and other fields without strong computational traditions. While the development of the portal is still in its early stages, this paper documents ongoing work towards an already operable pilot in addition to providing an overview of long-term goals and visions. At the core of the current pilot implementation we find Galaxy, a web-based workflow management system initially developed for data-intensive research in genomics and bioinformatics; therefore, an important part of the work on the pilot is to adapt and evaluate Galaxy for the context of a language analysis portal. Emanuele Lapponi, Erik Velldal, Nikolay A. Vazov, Stephan Oepen (2013). Towards Large-Scale Language Analysis in the Cloud, Proceedings of the workshop on Nordic language research infrastructure at NODALIDA 2013, May 22-24, 2013, Oslo, Norway. NEALT Proceedings Series 20 http://www.ep.liu.se/ecp_article/index.en.aspx?issue=089;article=001

We compare three different approaches to parsing into syntactic, bi-lexical dependencies for English: a ‘direct’ data-driven dependency parser, a statistical phrase structure parser, and a hybrid, ‘deep’ grammar-driven parser. The analyses from the latter two are post-converted to bilexical dependencies. Through this ‘reduction’ of all three approaches to syntactic dependency parsers, we determine empirically what performance can be obtained for a common set of dependency types for English, across a broad variety of domains. In doing so, we observe what trade-offs apply along three dimensions, accuracy, efficiency, and resilience to domain variation. Our results suggest that the hand-built grammar in one of our parsers helps in both accuracy and cross-domain performance. Proceedings of The 13th International Conference on Parsing Technologies IWPT-2013.

The Deep Linguistic Processing with HPSG Initiative (DELH-IN) provides the infrastructure needed to produce open-source semantic transfer-based machine translation systems. We have made available a prototype Japanese-English machine translation system built from existing resources include parsers, generators, bidirectional grammars and a transfer engine.

We discuss the use of NL systems in the domain of appointment scheduling. Appointment scheduling is a problem faced daily by many people and organizations, and typically solved using communication in natural language. In general, cooperative interaction between several participants is required whose calendar data are distributed rather than centralized. In this distributed multi-agent environment, the use of NL systems makes it possible for machines and humans to cooperate in solving scheduling problems. We describe the COSMA (Cooperative Schedule Managament Agent) system, a secretarial assistant for appointment scheduling. A central part of COSMA is the reusable NL core system DISCO, which serves, in this application, as an NL interface between an appointment planning system and the human user. COSMA is fully implemented in Common Lisp and runs on Unix Workstations. Our experience with COSMA shows that it is a plausible and useful application for NL systems. However, the appointment planner was not designed for NL communication and thus makes strong assumptions about sequencing of domain actions and about the error-freeness of the communication. We suggest that further improvements of the overall COSMA functionality, especially with regard to flexibility and robustness, be based on a modified architecture.

This paper recommends an approach to the implementation of semantic representation languages (SRLs) which exploits a parallelism between SRLs and programming languages (PLs). The design requirements of SRLs for natural language are similar to those of PLs in their goals. First, in both cases we seek modules in which both the surface representation (print form) and the underlying data structures are important. This requirement highlights the need for general tools allowing the printing and reading of expressions (data structures). Second, these modules need to cooperate with foreign modules, so that the importance of interface technology (compilation) is paramount; and third, both compilers and semantic modules need "inferential" facilities for transforming (simplifying) complex expressions in order to ease subsequent processing. But the most important parallel is the need in both fields for tools which are useful in combination with a variety of concrete languages -- general purpose parsers, printers, simplifiers (transformation facilities) and compilers. This arises in PL technology from (among other things) the need for experimentation in language design, which is again parallel to the case of SRLs. Using a compiler-based approach, we have implemented NLL, a public domain software package for computational natural language semantics. Several interfaces exist both for grammar modules and for applications, using a variety of interface technologies, including especially compilation. We review here a variety of NLL, applications, focusing on COSMA, an NL interface to a distributed appointment manager.

The Deep Linguistic Processing with HPSG Initiative (DELPH-IN) provides the infrastructure needed to produce open-source semantic transfer-based machine translation systems. We have made available a prototype Japanese-English machine translation system built from existing resources include parser, generators, bidirectional grammars and a transfer engine. Accepted version

In the LOGON machine translation system where semantic transfer using Minimal Recursion Semantics is being developed in conjunction with two existing broad-coverage grammars of Norwegian and English, we motivate the use of a grammar-specific semantic interface (SEM-I) to facilitate the construction and maintenance of a scalable translation engine. The SEM-I is a theoretically grounded component of each grammar, capturing several classes of lexical regularities while also serving the crucial engineering function of supplying a reliable and complete specification of the elementary predications the grammar can realize. We make extensive use of underspecification and type hierarchies to maximize generality and precision. Accepted version

The TSNLP project has investigated various aspects of the construction, maintenance and application of systematic test suites as diagnostic and evaluation tools for NLP applications. The paper summarizes the motivation and main results of the project: besides the solid methodological foundation, TSNLP has produced substantial multi-purpose and multi-user test suites for three European languages together with a set of specialized tools that facilitate the construction, extension, maintenance, retrieval, and customization of the test data. As TSNLP results, including the data and technology, are made publicly available, the project presents a valuable linguistic resourc e that has the potential of providing a wide-spread pre-standard diagnostic and evaluation tool for both developers and users of NLP applications., Comment: 7 pages, uses colap.sty and oe.sty. tar gzip uuencode. To appear in Proceedings of COLING-96

The German nominal group - even when reduced to the core inventory of nouns, determiners and attributive adjectives - is a morphologically and syntactically complex structure. In this paper it is suggested that a detailed understanding of the (morpho-) syntactic categories and the syntagmatic relations exhibited in the core nominal group is a prerequisite to an adequate analysis. It will be argued that the two fundamental syntagmatic relations holding within the nominal group, viz. GOVERNMENT and AGREEMENT, have to figure as theoretically primitive concepts in any reasonably detailed account of nominal structures. Explicating government and agreement relations and especially separating one from the other, will presuppose a sufficient inventory of formal descriptive devices in any particular theory of grammar. The paper is settled in the framework of Head-Driven Phrase Structure Grammar (HPSG). Recent HPSG analyses for the German nominal group that have been put forth in [Pollard and Sag 1994] and [Netter 1994] are studied in detail contrasting them to (semi-) formal proposals from other linguistic frameworks; potential problems as weIl as some abstract joint properties of the two HPSG approaches are exemplified. Building on this comparison it is concluded that in exactly the linguistic stipulations shared by the two accounts, two important generalizations about the inherent structure 'of the German nominal group are to be found. At the same time the [Pollard and Sag 1994] analysis is tentatively reformulated.

The natural language system DISCO is described. It combines o a powerful and flexible grammar development system; o linguistic competence for German including morphology, syntax and semantics; o new methods for linguistic performance modelling on the basis of high-level competence grammars; o new methods for modelling multi-agent dialogue competence; o an interesting sample application for appointment scheduling and calendar management.

Strong and affordable in-domain data is a desirable asset when transferring trained semantic parsers to novel domains. As previous methods for semi-automatically constructing such data cannot handle the complexity of realistic SQL queries, we propose to construct SQL queries via context-dependent sampling, and introduce the concept of topic. Along with our SQL query construction method, we propose a novel pipeline of semi-automatic Text-to-SQL dataset construction that covers the broad space of SQL queries. We show that the created dataset is comparable with expert annotation along multiple dimensions, and is capable of improving domain transfer performance for SOTA semantic parsers.

This paper shows the extent to which treebanks of Ancient Greek play a central role in the ongoing Pedalion project at the University of Leuven. Building on diverse treebanks readily available today, the project aims to make progress in the automated parsing of classical and postclassical Greek texts. Rather than developing new technology as such, our project endeavours to make deliberate and methodical use of the technology that already exists, essentially by combining and adapting both technology and data. This contribution offers a ‘roadmap’ of our project, surveying (a) the existing work on which we can rely, (b) the strategies which we adopt to reach better results in the automated processing of Ancient Greek and (c) the deliverables that have already been realised or are forthcoming. no ISSN ispartof: pages:109-117 ispartof: Proceedings of the 18th International Workshop on Treebanks and Linguistic Theories (TLT, SyntaxFest 2019) pages:109-117 ispartof: 18th International Workshop on Treebanks and Linguistic Theories (TLT, SyntaxFest 2019) location:Paris, France date:28 Aug - 29 Aug 2019 status: Published online

This paper presents experiments with document-level machine translation with readability constraints. We describe the task of producing simplified translations from a given source with the aim to optimize machine translation for specific target users such as language learners. In our approach, we introduce global features that are known to affect readability into a document-level SMT decoding framework. We show that the decoder is capable of incorporating those features and that we can influence the readability of the output as measured by common metrics. This study presents the first attempt of jointly performing machine translation and text simplification, which is demonstrated through the case of translating parliamentary texts from English to Swedish.

Although several syntactically annotated corpora (or treebanks) exist for Dutch, they are seldomly used for descriptive linguistic research because there are no easy-to-use exploitation tools available. This demonstration paper describes GrETEL, a linguistic search engine (http://nederbooms.ccl.kuleuven.be/eng/gretel) that enables non-technical users to consult treebanks in a user-friendly way. Instead of a formal search expression, a natural language example is used as input to the system, allowing users to search for similar constructions as the example they provide. In the first version of GrETEL, only written Dutch (LASSY) was included. Based on user requests we have now included the Spoken Dutch Corpus (CGN) as well. ispartof: pages:423-428 ispartof: Proceedings of the 19th Nordic Conference of Computational Linguistics (NODALIDA 2013) pages:423-428 ispartof: NoDaLiDa location:Oslo date:22 May - 24 May 2013 status: published

The RWAAI (Repository and Workspace for Austroasiatic Intangible heritage) project aims at building a digital archive out of existing legacy data from the austroasiatic language family. One aspect of the project is the preservation of analogue legacy data. In this context, we have at our hands a large number of mostly-phonemic transcriptions of narrative monologues, often with accompanying sound recordings, in the unwritten Kammu language of northern Laos. Some of the transcriptions, however, lack tone marks, which for a tonal language such as Kammu makes them substantially less useful. The problem of restoring tones can be recast as one of word sense disambiguation, or, more generally, lexical ambiguity resolution. We attack it by decision lists, along the lines of Yarowsky (1994), using the tone-marked part of the corpus (120kW) as training data. The performance ceiling of this corpus is uncertain: the stories were all annotated, primarily for human rather than machine consumption, by a single person during almost 40 years, with slowly emerging idiosyncratic conventions. Thus, both inter-annotator and intra-annotator agreement figures are unknown. Nevertheless, with the data from this one annotator as a gold standard, we improve from an already-high baseline accuracy of 95.7% to 97.2% (by 10-fold cross-validation).