Korpuslinguistik
Refine
Year of publication
- 2014 (22) (remove)
Document Type
- Conference Proceeding (13)
- Part of a Book (5)
- Article (3)
- Book (1)
Has Fulltext
- yes (22)
Keywords
Publicationstate
Reviewstate
- (Verlags)-Lektorat (7)
- Peer-Review (2)
Publisher
- European Language Resources Association (ELRA) (5)
- Institut für Deutsche Sprache (4)
- ELRA (2)
- European Language Resources Association (2)
- Association for Computational Linguistics and Dublin City University (1)
- Institute for Specialised Communication and Multilingualism (1)
- Narr (1)
- Universität Hildesheim (1)
- Universitätsverlag Hildesheim (1)
- de Gruyter (1)
We investigate how the granularity of POS tags influences POS tagging, and furthermore, how POS tagging performance relates to parsing results. For this, we use the standard “pipeline” approach, in which a parser builds its output on previously tagged input. The experiments are performed on two German treebanks, using three POS tagsets of different granularity, and six different POS taggers, together with the Berkeley parser. Our findings show that less granularity of the POS tagset leads to better tagging results. However, both too coarse-grained and too fine-grained distinctions on POS level decrease parsing performance.
We discovered several recurring errors in the current version of the Europarl Corpus originating both from the web site of the European Parliament and the corpus compilation based thereon. The most frequent error was incompletely extracted metadata leaving non-textual fragments within the textual parts of the corpus files. This is, on average, the case for every second speaker change. We not only cleaned the Europarl Corpus by correcting several kinds of errors, but also aligned the speakers’ contributions of all available languages and compiled every- thing into a new XML-structured corpus. This facilitates a more sophisticated selection of data, e.g. querying the corpus for speeches by speakers of a particular political group or in particular language combinations.
Designing a Bilingual Speech Corpus for French and German Language Learners: a Two-Step Process
(2014)
We present the design of a corpus of native and non-native speech for the language pair French-German, with a special emphasis on phonetic and prosodic aspects. To our knowledge there is no suitable corpus, in terms of size and coverage, currently available for the target language pair. To select the target L1-L2 interference phenomena we prepare a small preliminary corpus (corpus1), which is analyzed for coverage and cross-checked jointly by French and German experts. Based on this analysis, target phenomena on the phonetic and phonological level are selected on the basis of the expected degree of deviation from the native performance and the frequency of occurrence. 14 speakers performed both L2 (either French or German) and L1 material (either German or French). This allowed us to test, recordings duration, recordings material, the performance of our automatic aligner software. Then, we built corpus2 taking into account what we learned about corpus1. The aims are the same but we adapted speech material to avoid too long recording sessions. 100 speakers will be recorded. The corpus (corpus1 and corpus2) will be prepared as a searchable database, available for the scientific community after completion of the project.
This paper presents the first release of the KiezDeutsch Korpus (KiDKo), a new language resource with multiparty spoken dialogues of Kiezdeutsch, a newly emerging language variety spoken by adolescents from multi-ethnic urban areas in Germany. The first release of the corpus includes the transcriptions of the data as well as a normalisation layer and part-of-speech annotations. In the paper, we describe the main features of the new resource and then focus on automatic POS tagging of informal spoken language. Our tagger achieves an accuracy of nearly 97% on KiDKo. While we did not succeed in further improving the tagger using ensemble tagging, we present our approach to using the tagger ensembles for identifying error patterns in the automatically tagged data.
Der korpuslinguistische Ansatz des Projekts »Korpusgrammatik« eröffnet neue Perspektiven auf unsere Sprachwirklichkeit allgemein und grammatische Regularitäten im Besonderen. Der vorliegende Band klärt auf, wie man korpuslinguistisch nach dem Standard fragen kann, wie die Projektkorpora aufgebaut und in einer Korpusdatenbank erschlossen sind, wie man in einem automatischen Abfragesystem der Variabilität der Sprache zu Leibe rückt und sie sogar messbar macht, schließlich aber auch, wo die Grenzen quantitativer Korpusanalysen liegen. Pilotstudien deuten an, wie der Ansatz unsere grammatischen Horizonte erweitert und die Grammatikografie voranbringt.
Dieser Beitrag stellt das Forschungs- und Lehrkorpus Gesprochenes Deutsch (FOLK) und die Datenbank für Gesprochenes Deutsch (DGD) als Instrumente gesprächsanalytischer Arbeit vor. Nach einer allgemeinen Einführung in FOLK und DGD im zweiten Abschnitt werden im dritten Abschnitt die methodischen Beziehungen zwischen Korpuslinguistik und Gesprächsforschung und die Herausforde-rungen, die sich bei der Begegnung dieser beiden Herangehensweisen an authenti-sches Sprachmaterial stellen, kurz skizziert. Der vierte Abschnitt illustriert dann ausgehend vom Beispiel der Formel ich sag mal, wie eine korpus- und datenbankgesteuerte Analyse zur Untersuchung von Gesprächsphänomenen beitragen kann.
Wikipedia is a valuable resource, useful as a lingustic corpus or a dataset for many kinds of research. We built corpora from Wikipedia articles and talk pages in the I5 format, a TEI customisation used in the German Reference Corpus (Deutsches Referenzkorpus - DeReKo). Our approach is a two-stage conversion combining parsing using the Sweble parser, and transformation using XSLT stylesheets. The conversion approach is able to successfully generate rich and valid corpora regardless of languages. We also introduce a method to segment user contributions in talk pages into postings.
Part-of-speech tagging (POS-tagging) of spoken data requires different means of annotation than POS-tagging of written and edited texts. In order to capture the features of German spoken language, a distinct tagset is needed to respond to the kinds of elements which only occur in speech. In order to create such a coherent tagset the most prominent phenomena of spoken language need to be analyzed, especially with respect to how they differ from written language. First evaluations have shown that the most prominent cause (over 50%) of errors in the existing automatized POS-tagging of transcripts of spoken German with the Stuttgart Tübingen Tagset (STTS) and the treetagger was the inaccurate interpretation of speech particles. One reason for this is that this class of words is virtually absent from the current STTS. This paper proposes a recategorization of the STTS in the field of speech particles based on distributional factors rather than semantics. The ultimate aim is to create a comprehensive reference corpus of spoken German data for the global research community. It is imperative that all phenomena are reliably recorded in future part-of-speech tag labels.
Machine learning methods offer a great potential to automatically investigate large amounts of data in the humanities. Our contribution to the workshop reports about ongoing work in the BMBF project KobRA (http://www.kobra.tu-dortmund.de) where we apply machine learning methods to the analysis of big corpora in language-focused research of computer-mediated communication (CMC). At the workshop, we will discuss first results from training a Support Vector Machine (SVM) for the classification of selected linguistic features in talk pages of the German Wikipedia corpus in DeReKo provided by the IDS Mannheim. We will investigate different representations of the data to integrate complex syntactic and semantic information for the SVM. The results shall foster both corpus-based research of CMC and the annotation of linguistic features in CMC corpora.
Maximizing the potential of very large corpora: 50 years of big language data at IDS Mannheim
(2014)
Very large corpora have been built and used at the IDS since its foundation in 1964. They have been made available on the Internet since the beginning of the 90’s to currently over 30,000 researchers worldwide. The Institute provides the largest archive of written German (Deutsches Referenzkorpus, DeReKe) which has recently been extended to 24 billion words. DeReKe has been managed and analysed by engines known as COSMAS and afterwards COSMAS II, which is currently being replaced by a new, scalable analysis platform called KorAP. KorAP makes it possible to manage and analyse texts that are accompanied by multiple, potentially conflicting, grammatical and structural annotation layers, and is able to handle resources that are distributed across different, and possibly geographically distant, storage systems. The majority of texts in DeReKe are not licensed for free redistribution, hence, the COSMAS and KorAP systems offer technical solutions to facilitate research on very large corpora that are not available (and not suitable) for download. For the new KorAP system, it is also planned to provide sandboxed environments to support non-remote-API access “near the data” through which users can run their own analysis programs.
We describe a systematic and application-oriented approach to training and evaluating named entity recognition and classification (NERC) systems, the purpose of which is to identify an optimal system and to train an optimal model for named entity tagging DeReKo, a very large general-purpose corpus of contemporary German (Kupietz et al., 2010). DeReKo 's strong dispersion wrt. genre, register and time forces us to base our decision for a specific NERC system on an evaluation performed on a representative sample of DeReKo instead of performance figures that have been reported for the individual NERC systems when evaluated on more uniform and less diverse data. We create and manually annotate such a representative sample as evaluation data for three different NERC systems, for each of which various models are learnt on multiple training data. The proposed sampling method can be viewed as a generally applicable method for sampling evaluation data from an unbalanced target corpus for any sort of natural language processing.
We present an approach to an aspect of managing complex access scenarios to large and heterogeneous corpora that involves handling user queries that, intentionally or due to the complexity of the queried resource, target texts or annotations outside of the given user’s permissions. We first outline the overall architecture of the corpus analysis platform KorAP, devoting some attention to the way in which it handles multiple query languages, by implementing ISO CQLF (Corpus Query Lingua Franca), which in turn constitutes a component crucial for the functionality discussed here. Next, we look at query rewriting as it is used by KorAP and zoom in on one kind of this procedure, namely the rewriting of queries that is forced by data access restrictions.
We start by trying to answer a question that has already been asked by de Schryver et al. (2006): Do dictionary users (frequently) look up words that are frequent in a corpus. Contrary to their results, our results that are based on the analysis of log files from two different online dictionaries indicate that users indeed look up frequent words frequently. When combining frequency information from the Mannheim German Reference Corpus and information about the number of visits in the Digital Dictionary of the German Language as well as the German language edition of Wiktionary, a clear connection between corpus and look-up frequencies can be observed. In a follow-up study, we show that another important factor for the look-up frequency of a word is its temporal social relevance. To make this effect visible, we propose a de-trending method where we control both frequency effects and overall look-up trends.
Der Beitrag beschäftigt sich mit der Frage, wie und inwieweit korpusbasierte Ansätze zur Untersuchung und Bewertung von Sprachwandel beitragen können. Die Bewertung von Sprachwandel erscheint in dieser Hinsicht interessant, da sie erstens von größerem öffentlichen Interesse ist, zweitens nicht zu den Kernthemen der Sprachwissenschaft zählt und drittens sowohl die geisteswissenschaftlichen Aspekte der Sprachwissenschaft berührt als auch die empirischen, die eher für die so genannten harten Wissenschaften typisch sind. Letzteres trifft bei der Frage nach Sprachverfall (gutem vs. schlechtem Deutsch diachron) vermutlich unbestrittener zu als bei der Frage nach richtigem vs. falschem Deutsch, da zu ihrer Beantwortung offensichtlich einerseits empirische, messbare Kriterien herangezogen werden müssen, andererseits aber auch weitere Kriterien notwendig sind und es außerdem einer Entscheidung zur Einordnung und Gewichtung der verschiedenartigen Kriterien sowie einer Begründung dieser Entscheidung bedarf. Zur Annäherung an die Fragestellung werden zunächst gängige, leicht operationalisierbare Hypothesen zu Symptomen eines potenziellen Verfalls des Deutschen auf verschiedenen DeReKo-basierten Korpora überprüft und im Hinblick auf ihre Verallgemeinerbarkeit und Tragweite diskutiert. Im zweiten Teil werden weitere empirische Ansätze zur Untersuchung von Wandel, Variation und Dynamik skizziert, die zur Diskussion spezieller Aspekte von Sprachverfall beitragen könnten. Im Schlussteil werden die vorgestellten Ansätze in den Gesamtkontext einer sprachwissenschaftlichen Untersuchung von Sprachverfall gestellt und vor dem Hintergrund seines gesellschaftlichen Diskurses reflektiert.
Language resources are often compiled for the purpose of variational analysis, such as studying differences between genres, registers, and disciplines, regional and diachronic variation, influence of gender, cultural context, etc. Often the sheer number of potentially interesting contrastive pairs can get overwhelming due to the combinatorial explosion of possible combinations. In this paper, we present an approach that combines well understood techniques for visualization heatmaps and word clouds with intuitive paradigms for exploration drill down and side by side comparison to facilitate the analysis of language variation in such highly combinatorial situations. Heatmaps assist in analyzing the overall pattern of variation in a corpus, and word clouds allow for inspecting variation at the level of words.
Data Mining with Shallow vs. Linguistic Features to Study Diversification of Scientific Registers
(2014)
We present a methodology to analyze the linguistic evolution of scientific registers with data mining techniques, comparing the insights gained from shallow vs. linguistic features. The focus is on selected scientific disciplines at the boundaries to computer science (computational linguistics, bioinformatics, digital construction, microelectronics). The data basis is the English Scientific Text Corpus (SCITEX) which covers a time range of roughly thirty years (1970/80s to early 2000s) (Degaetano-Ortlieb et al., 2013; Teich and Fankhauser, 2010). In particular, we investigate the diversification of scientific registers over time. Our theoretical basis is Systemic Functional Linguistics (SFL) and its specific incarnation of register theory (Halliday and Hasan, 1985). In terms of methods, we combine corpus-based methods of feature extraction and data mining techniques.
Newspapers became extremely popular in Germany during the 18th and 19th century, and thus increasingly influential for modern German. However, due to the lack of digitized historical newspaper corpora for German, this influence could not be analyzed systematically. In this paper, we introduce the Mannheim Corpus of Digital Newspapers and Magazines, which in its current release comprises 21 newspapers and magazines from the 18th and 19th century. With over 4.1 Mio tokens in about 650 volumes it currently constitutes the largest historical corpus dedicated to newspapers in German. We briefly discuss the prospect of the corpus for analyzing the evolution of news as a genre in its own right and the influence of contextual parameters such as region and register on the language of news. We then focus on one historically influential aspect of newspapers – their role in disseminating foreign words in German. Our preliminary quantitative results indeed indicate that newspapers use foreign words significantly more frequently than other genres, in particular belles lettres.