Introducing, evaluating ukWaC, a very large web-derived corpus of English

By Adriano Ferraresi, Eros Zanchetta, Marco Baroni, Silvia Bernardini. 2008

328 citations recorded by the source corpus.

Subfields: Natural Language Processing, Evaluation & Benchmarks

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Natural Language Processing

In Evaluation & Benchmarks

Citations in the corpus

context2vec: Learning Generic Context Embedding with Bidirectional LSTM (2016, methodology): “With the exception of the sentence completion task (MSCC), which comes with its own learning corpus, we used the two billion word ukWaC (Ferraresi et al., 2008) as our learning corpus.”
Improving Distributional Similarity with Lessons Learned from Word Embeddings (2015, methodology): “This corpus was built from an 8.5 billion word corpus sug-219 gested by Mikolov for training word2vec , 7 to which we added UKWaC (Ferraresi et al., 2008).”
Evaluating Neural Word Representations in Tensor-Based Compositional Settings (2014, methodology): “In this variation, we train a vector space from the ukWaC corpus 3 (Ferraresi et al., 2008), originally using as a basis the 2,000 content words with the highest frequency (but excluding a list of stop words as well as the 50 most frequent content words since they exhibit low information content).”
A New Dataset and Method for Automatically Grading ESOL Texts (2011, methodology): “In order to estimate the error-rate, we build a trigram language model (LM) using ukWaC (ukWaC LM) (Ferraresi et al., 2008), a large corpus of English containing more than 2 billion tokens.”
Making Sense of Word Embeddings (2016, methodology): “For experiments, we use two commonly used corpora for training distributional models: Wikipedia2 and ukWaC (Ferraresi et al., 2008).”
Novel Word-sense Identification (2014, methodology): “In this study we consider two datasets of relatively newly-coined word-senses: (1) an extended version of the dataset based on the BNC (Burnard, 2000) and ukWaC (Ferraresi et al., 2008) used by Lau et al.”
DEFT: A corpus for definition extraction in free- and semi-structured text (2019, methodology): “…extraction has been applied solely to English datasets, including the WCL dataset mentioned above (Navigli et al., 2010), the ukWaC dataset (Ferraresi et al., 2008), a large crawled dataset of the .uk domain name, and the W00 dataset, a small, expertly annotated corpus introduced by Jin…”
Modeling Word Meaning in Context with Substitute Vectors (2015, methodology): “All compared methods are unsupervised and use the plain text of ukWaC (Ferraresi et al., 2008), a two 3 Our pseudo-word dataset is available at: www.cs.biu. ac.il/nlp/resources/downloads/word2parvec/ billion word web corpus, as their learning corpus.”
Human and Machine Judgements for Russian Semantic Relatedness (2016, methodology): “The creators of the MEN dataset [22] went even further, annotating via crowdsourcing 3 000 word pairs sampled from the ukWaC corpus [23].”
Annotated Gigaword (2012, methodology): “1 The ukWaC corpus was developed by crawling the .uk domain, resulting in nearly 2-billion words then annotated with part-of-speech tags and lemmas (Ferraresi et al., 2008). ukWaC was later extended to include dependency parses extracted using the MaltParser (Nivre et al., 2007) (PukWaC).”
Word Sense Induction for Novel Sense Detection (2012, methodology): “In particular, we take the written portion of the BNC — consisting primarily of British English text from the late 20th century — as our reference corpus, and a similarlysized random sample of documents from the ukWaC (Ferraresi et al., 2008) — a Web corpus built from the .”
Have You Lost the Thread? Discovering Ongoing Conversations in Scattered Dialog Blocks (2017, methodology): “We extracted distributional vectors from the ukWaC corpus (Ferraresi et al. 2008) with the following settings of word2vec : min-count = 50 , window = 5 , iter = 10 , negative = 10 , and vector size = 250 .”
A Simple Word Embedding Model for Lexical Substitution (2015, methodology): “We used ukWaC (Ferraresi et al., 2008), a two billion word web corpus, as our learning corpus.”
Boilerplate detection using shallow text features (2010, background): “Another CleanEval contestant, BTE, determines the largest contiguous text area with the least amount of HTML tags and marks it as “full text” [12,13].”
Prior Disambiguation of Word Tensors for Constructing Sentence Vectors (2013, methodology): “Experimental setting We train our vectors using ukWaC (Ferraresi et al., 2008), a corpus of English text with 2 billion words (100m sentences).”
Compositional Operators in Distributional Semantics (2014, positive): “…have been widely studied and successfully applied on a variety of language tasks, especially during the last decade with the availability of large-scale corpora, like Gigaword (Graff et al., 2003) and ukWaC (Ferraresi et al., 2008), which provide a reliable resource for training the vector spaces.”
SemEval-2020 Task 6: Definition Extraction from Free Text with the DEFT Corpus (2020, background): “Given the precedence of previous performance on other more constrained relation extraction tasks (see also Lin et al. (2016), Shen and Huang (2016), Sorokin and Gurevych (2017), etc.), we predicted this task would see mostly low performing scores.”
Detecting Asymmetric Semantic Relations in Context: A Case-Study on Hypernymy Detection (2017, methodology): “Contextualized Word Representations To obtain the Context2Vec representations, we use an existing 600-dimensional model trained on ukWaC (Ferraresi et al., 2006).”
XHate-999: Analyzing and Detecting Abusive Language Across Domains and Languages (2020, positive): “In short, we train a logistic regression classiﬁer on each training set separately, rank the words according to the weights associated with the abusive class, and retain only the ones which occur in the top 10k most frequent English words in the ukWaC corpus (Ferraresi et al., 2008).”