Matches in ScholarlyData for { ?s ?p ?o. }
- 2 abstract "For the acquisition of common-sense knowledge as well as as a way to answer linguistic questions regarding actual language usage, the breadth and depth of the World Wide Web has been welcomed to supplement large text corpora (usually from newspapers) as a useful resource. Previous research using Web corpora for either knowledge acquisition or information extraction has sometimes shown them to be less useful than newspaper or newswire corpora. More than a-priori criticism from corpus linguists on the difficulty of assuring balanced text composition and/or text quality does, these empirical results underline the importance of assessing and improving the quality of Web corpora to ensure their usefulness in real-world tasks. In this paper, we present our own pipeline for Web corpora, which includes improvements regarding content-sensitive boilerplate detection as well as language filtering for mixed-language documents, and provide a task-based evaluation of the combination of corpora and (non-linguistic and linguistic) preprocessing between more standard types of large corpora (newspaper and Wikipedia) and different Web corpora. While our current results are focused on German-language Web corpora, both the content-sensitive boilerplate detection and our method of evaluation by constructing an artificial thesaurus from a wordnet are applicable to many other languages.".
- 2 hasAuthorList authorList.
- 2 isPartOf proceedings.
- 2 keyword "boilerplate removal".
- 2 keyword "corpus building".
- 2 keyword "evaluation".
- 2 title "Not Just Bigger: Towards Better-Quality Web Corpora".
- authorList type List.
- authorList hasFirstItem item-1.
- authorList hasItem item-1.
- authorList hasItem item-2.
- authorList hasLastItem item-2.
- item-1 type ListItem.
- item-1 hasContent yannick-versley.
- item-1 next item-2.
- item-2 type ListItem.
- item-2 hasContent yana-panchenko.
- item-2 previous item-1.
- 3 creator marco-brunello.
- 3 type InProceedings.
- 3 label "Understanding the composition of parallel corpora from the web".
- 3 sameAs 3.
- 3 abstract "Although it is fundamental to have a good fit between the text typology of training data and to-be-translated data in machine translation, there is a lack of studies on analysing parallel data under this point of view. This paper describes some studies made with the aim of understanding the composition of parallel corpora, in particular by using topic modeling.".
- 3 hasAuthorList authorList.
- 3 isPartOf proceedings.
- 3 keyword "camera-ready version".
- 3 keyword "machine translation".
- 3 keyword "parallel corpora".
- 3 keyword "topic modeling".
- 3 keyword "web as corpus".
- 3 title "Understanding the composition of parallel corpora from the web".
- authorList type List.
- authorList hasFirstItem item-1.
- authorList hasItem item-1.
- authorList hasLastItem item-1.
- item-1 type ListItem.
- item-1 hasContent marco-brunello.
- 4 creator jan-pomikalek.
- 4 creator vit-suchomel.
- 4 type InProceedings.
- 4 label "Efficient Web Crawling for Large Text Corpora".
- 4 sameAs 4.
- 4 abstract "Many researchers use texts from the web, an easy source of linguistic data in a great variety of languages. Building both large and good quality text corpora is the challenge we face nowadays. In this paper we describe how to deal with inefficient data downloading and how to focus crawling on text rich web domains. The idea has been successfully implemented in SpiderLing. We present efficiency figures from crawling texts in American Spanish, Czech, Japanese, Russian, Tajik Persian, Turkish and the sizes of the resulting corpora.".
- 4 hasAuthorList authorList.
- 4 isPartOf proceedings.
- 4 keyword "corpus".
- 4 keyword "crawler".
- 4 keyword "text corpus".
- 4 keyword "web corpus".
- 4 keyword "web crawling".
- 4 title "Efficient Web Crawling for Large Text Corpora".
- authorList type List.
- authorList hasFirstItem item-1.
- authorList hasItem item-1.
- authorList hasItem item-2.
- authorList hasLastItem item-2.
- item-1 type ListItem.
- item-1 hasContent vit-suchomel.
- item-1 next item-2.
- item-2 type ListItem.
- item-2 hasContent jan-pomikalek.
- item-2 previous item-1.
- 5 creator tobias-roth.
- 5 type InProceedings.
- 5 label "Using Web Corpora for the Recognition of Regional Variation in Standard German Collocations".
- 5 sameAs 5.
- 5 abstract "The Standard German collocation "berührende Worte" (touching words) is only used in Austria. Yet, it consists of exclusively Common German component words. It is the combination that makes it a regional variant - one on a purely collocational level. One of the goals in our German collocations dictionary project is to describe regional variation in the collocations collected. The main tool for this is a web corpus divided into three subcorpora. Each of them contains web pages from a different country code top-level domain (Austria, Germany and Switzerland). Our method of variation recognition using a web corpus is compared to previously used resources. Advantages, disadvantages and results of the web-corpus approach are discussed. Given that the extent of regional variation in German collocations and its structuring into collocational, lexical and syntactical variation has been unknown, we present first estimates in that respect.".
- 5 hasAuthorList authorList.
- 5 isPartOf proceedings.
- 5 keyword "German language".
- 5 keyword "collocations".
- 5 keyword "corpus linguistics".
- 5 keyword "regional variation".
- 5 title "Using Web Corpora for the Recognition of Regional Variation in Standard German Collocations".
- authorList type List.
- authorList hasFirstItem item-1.
- authorList hasItem item-1.
- authorList hasLastItem item-1.
- item-1 type ListItem.
- item-1 hasContent tobias-roth.
- 6 creator dayne-freitag.
- 6 creator ed-chow.
- 6 creator john-niekrasz.
- 6 creator paul-kalmar.
- 6 creator tulay-muezzinoglu.
- 6 type InProceedings.
- 6 label "A corpus of online discussions for research into linguistic memes".
- 6 sameAs 6.
- 6 abstract "We describe a 460-million word corpus of online discussions. The data are collected from public news websites and community-of-interest Internet forums, and are designed to support research on the propagation of socially relevant ideas, a.k.a., "memes." A structural and statistical description of the corpus is given, and the employed methods of website monitoring, collection, and extraction are described. We also present preliminary linguistic research on the corpus. We show that the corpus represents language from a wide variety of social and psychological communities, that discussion structure and popularity can be predicted in large part from lexical analysis, and that standard epidemiological models provide good fit for diachronic patterns of population-level lexical adoption.".
- 6 hasAuthorList authorList.
- 6 isPartOf proceedings.
- 6 keyword "information diffusion".
- 6 keyword "memes".
- 6 keyword "social media".
- 6 keyword "socio-linguistics".
- 6 title "A corpus of online discussions for research into linguistic memes".
- authorList type List.
- authorList hasFirstItem item-1.
- authorList hasItem item-1.
- authorList hasItem item-2.