Journal for Language Technology and Computational Linguistics (JLCL)

Pölitz, Christian

English

Data Mining Software for Corpus Linguistics with Application in Diachronic Linguistics

Lemnitzer, Lothar

Editorial 
Ich begrüße Sie zur dieser thematisch nicht gebundenen Ausgabe des 
Journals.
Es ist das erste (und einzige) Heft des 30. Jahrgangs (ein Jubiläumsjahrgang, wenn Sie so wollen). 
Das Heft ist das Ergebnis einer Ausschreibung und der anschließenden 
Begutachtung der Einreichungen, für die ich mich herzlich bei allen GutachterInnen 
bedanken möchte.
Nach dem Begutachtungsprozess blieben fünf Beiträge übrig, die in diesem Heft 
zusammengestellt sind. In diesem Zusammenhang gilt mein Dank auch den Autorinnen und 
Autoren, die ihre Zeit und Mühe auf die von den Gutachtern vorgeschlagenen Überarbeitungen 
aufgewendet haben.
Amine Bayoudhi und Kollegen nehmen ein „multi-domain“ Korpus des Arabischen zum 
Ausgangspunkt für die Annoation der (emotionalen) Gerichtetheit von Ausagen auf der Ebene von
Sätzen und Teilsätzen.
Alexander Magidow stellt in seinem Beitrag ein relationales Datenbankmodell für die Speicherung
verschiedenartiger Typen von linguistischen Daten vor. Auch in seiner Arbeit spielt die arabische 
Sprache eine wichtige Rolle.
Laurent Romary macht in seinem Beitrag einen Vorschlag, Details des sog. „Lexical Markup 
Framwork“ (LMF) in TEI zu serialisieren. Eines seiner Beispiele ist die Modellierung eines 
koreanischen Wortnetzes.
Uladzimir Sidarenka und Kollegen schlagen ein Verfahren für die Segmentierung von Texten eines 
deutschen Korpus in Diskurseinheiten vor und evaluieren dieses Verfahren.
Lucie Flekova und Kollegen nehmen Transkripte audivisueller Aufnahmen aus Unterrichtssituation 
zum Ausgangspunkt für eine Klassifizierung nach diskurspragmatischen Kategorien. Hier kommen
Verfahren des maschinellen Lernens zum Einsatz.
Zum Schluss noch ein Ausblick auf das Jahr 2016.  Für dieses Jahr sind wieder 2 Hefte geplant, die
jeweils auf Workshops basieren und „Korpora“ in verschiedenen Facetten  zum Thema haben 
werden.
Ich wünsche Ihnen aber zunächst einmal viel Spaß und intellektuellen Gewinn bei der Lektüre der 
Beiträge dieses Heftes.
Beste Grüße aus Berlin
Lothar Lemnitzer Januar 2016


Editorial

Magidow, Alexander

A relational database model and prototype for storing diverse discrete linguistic data

Bayoudhi, Amine

Ghorbel, Hatem

Koubaa, Houssem

Hadrich Belguith, Lamia

Sentiment Classification at Discourse Segment Level: Experiments on multi-domain Arabic corpus

Flekova, Lucie

Sousa, Tahir

Mieskes, Margot

Gurevych, Iryna

Document-level school lesson quality classification based on German transcripts

Romary, Laurent

TEI and LMF crosswalks

Sidarenka, Uladzimir

Peldszus, Andreas

Stede, Manfred

Discourse Segmentation of German Texts

van Halteren, Hans

Oostdijk, Nelleke

Variability in Dutch Tweets. An estimate of the proportion of deviant word tokens

Wiegand, Michael

Remus, Robert

Gindl, Stefan

iii 
 
 
 
Editorial 
  
  
The abundance of opinions available on the World Wide Web represents an information 
repository of enormous intellectual and economic value. Automated methods to exploit this 
rich knowledge mine have become more and more relevant within the last decade and the 
availability of large amounts of data is an ideal premise for the application of empirical 
methods. 
Although many researchers from different nations and institutes intensively work on the 
development of these techniques, many challenges have been left uncovered. The most 
pressing problems range from migrating sentiment analysis systems to new text types or 
domains, developing robust natural language applications that effectively exploit sentiment 
analysis, to the creation of resources that enable research in other languages than English. 
Moreover, a deeper understanding of subjective language beyond lexical keyword matching 
still needs to be acquired. 
This special issue consists of a selection of papers presented at the 2nd Workshop on 
Practice and Theory of Opinion Mining and Sentiment Analysis (PATHOS) held in conjunc-
tion with GSCL-2013 in Darmstadt, Germany, on September 23rd, 2013. In order to ensure 
articles of a high quality, a second reviewing cycle was carried out on the revised submis-
sions originally accepted and presented at the PATHOS workshop. 
We briefly outline the topics addressed in those papers: 
 
 Albertini et al. present an unsupervised method based on Growing Hierarchical 
Self-Organizing Maps to provide an alternative feature encoding. The aim of this 
encoding is to obtain a less sparse feature representation that typically arises with 
(traditional) bag of words applied on short documents. In the light of the growing 
importance of analyzing short texts from microblogging services, most prominent-
ly messages from Twitter, the task addressed by the authors is highly relevant to 
sentiment analysis. Their proposed encoding is evaluated against other competing 
methods (such as Autoencoders) and shown to outperform them. 
 
 Another paper that focuses on learning-based methods is Marchand et al. who ex-
amine multi-polarity words, i.e. polar expressions that change their polarity across 
different domains. As the set of domains on which sentiment analysis can be ap-
plied is pretty large, learning-based approaches often face the problem that only 
labeled out-of-domain training data are available. Marchand et al. show that the 
deletion of multi-polarity words substantially improves classification performance 
when such training data are used and propose a method to detect such words. They 
assume a realistic setting in which no labeled information from the target domain 
is available.  
 
 
 
 
iv 
 
 Ruppenhofer et al. describe the shared task on source and target extraction from 
political speeches which is to be organized in summer 2014. This article makes a 
welcome contribution to JLCL, being the flagship journal for research in German 
speaking countries, since it describes the first shared task that is exclusively con-
cerned with sentiment analysis in German. 
 
 Another work that focuses on sentiment analysis on a language other than English 
is presented by Veselovska et al. who introduce a subjectivity lexicon for Czech. 
The work describes the creation of the resource and its evaluation on polarity clas-
sification in four different domains and is an important example of resource crea-
tion for Czech.  
 
 Degaetano-Ortlieb et al. report on a study in a rather different direction. It is a de-
scriptive approach on sentiment analysis whose purpose is to uncover evaluative 
expressions with a focus on the notion of "importance" in the genre of scientific 
research articles. The study is carried out on a specially annotated corpus that al-
lows an examination of very complex linguistic properties. We believe that such 
research enables a deeper understanding of sentiment and subjective language than 
can be gained by the predominant resources, such as textual corpora labeled for 
polarity and sentiment lexicons.  
 
 The last article of this issue comes in a similar vein. Gu et al. present an explorato-
ry study of using electroencephalography (EEG) for the prediction of lexical va-
lence. This is a highly interdisciplinary work as it departs from traditional uni-
modal approaches of sentiment analysis that exclusively draw information for pre-
diction from text. This work is an example of the emerging research area of multi-
modal analysis that has recently attracted wide attention in sentiment analysis. 
 
We would like to thank the “Gesellschaft für Sprachtechnologie und Computerlinguistik” 
for accepting this volume to be published in the JLCL series. Special thanks also go to 
Thierry Declerck, the editor-in-chief of the journal, for supporting us with technical issues in 
creating this special issue. Moreover, we are indebted to our reviewers for their hard work. 
Last but not least, we thank the authors of the articles for their interesting contributions. 
 
August 2014 
 
Stefan Gindl, Robert Remus, Michael Wiegand 


Ruppenhofer, Josef

Struß, Julia Maria

Sonntag, Jonathan

IGGSA-STEPS: Shared Task on Source and Target Extraction from Political Speeches

Not a member yet

Data Mining Software for Corpus Linguistics with Application in Diachronic Linguistics

Editorial

A relational database model and prototype for storing diverse discrete linguistic data

Sentiment Classification at Discourse Segment Level: Experiments on multi-domain Arabic corpus

Document-level school lesson quality classification based on German transcripts

TEI and LMF crosswalks

Discourse Segmentation of German Texts

Variability in Dutch Tweets. An estimate of the proportion of deviant word tokens

Editorial

IGGSA-STEPS: Shared Task on Source and Target Extraction from Political Speeches

231

253