What to do about bad language on the internet

By Jacob Eisenstein. 2013

418 citations recorded by the source corpus.

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

Citations in the corpus

Language (Technology) is Power: A Critical Survey of “Bias” in NLP (2020, background): “For example, linguistic phenomena arising from many language practices (Eisenstein, 2013) are described as “noisy text” and often viewed as a target for “normalization.””
Honeycrisp: large-scale differentially private aggregation without a trusted core (2019, neutral): “5% over an entire year [30].”
Survey of the State of the Art in Natural Language Generation: Core tasks, applications and evaluation (2017, background): “However, social media data has given more impetus to the exploration of non-canonical language (e.g. Eisenstein, 2013); the impact of social and demographic factors on language use (e.g. Hovy & Søgaard, 2015; Johannsen et al., 2015); the prevalence of paralinguistic features such as affect, irony…”
Demographic Factors Improve Classification Performance (2015, background): “Also, rather than correlating with census-level statistics, as in (Eisenstein et al., 2011; Eisenstein, 2013a; Eisenstein, to appear), we take individual information of each author into account.”
Predicting crime using Twitter and kernel density estimation (2014, neutral): “, word boundary identification) [6].”
BERTweet: A pre-trained language model for English Tweets (2020, background): “…generally different from those of traditional written text such as Wikipedia and news articles, due to the typical short length of Tweets and frequent use of informal grammar as well as irregular vocabulary e.g. abbreviations, ty-pographical errors and hashtags (Eisenstein, 2013; Han et al., 2013).”
Cross-lingual syntactic variation over age and gender (2015, background): “Indeed, results show that this resource reﬂects the phonological and morpho-lexical variation of spoken language (Eisenstein, 2013b; Eisenstein, 2013a; Doyle, 2014).”
Emoticons vs. Emojis on Twitter: A Causal Inference Approach (2015, background): “Nonetheless, the overwhelming majority of Twitter messages are not near the character limit (Eisenstein 2013), indicating that this is unlikely to be the main reason for the decrease in emoticon characters — rather, it seems more likely that emojis are re-placing emoticons in fulﬁlling the same…”
Returning the N to NLP: Towards Contextually Personalized Classification Models (2020, background): “…treating demographic variables as ﬂuid and social, e.g. modeling what inﬂuences speakers to show more or less of their identity through language, or jointly modeling variation between and within speakers (Eckert and McConnell-Ginet, 2013; Nguyen et al., 2014; Bamman et al., 2014; Eisenstein, 2013).”
Whose Language Counts as High Quality? Measuring Language Ideologies in Text Data Selection (2022, background): “Language 572 on the Internet has been historically treated as a 573 noisy variant of English, even though lexical vari574 ation on the Internet is highly communicative of 575 social signals (Eisenstein, 2013), and varies con576 siderably along demographic variables (Eisenstein 577 et al.”
Unsupervised Domain Adaptation of Contextualized Embeddings for Sequence Labeling (2019, background): “Twitter poses some of the same challenges as historical text: orthographic variation leads to a substantial mismatch in vocabulary between the target domain and source training documents such as Wikipedia (Baldwin et al., 2013; Eisenstein, 2013).”
Five sources of bias in natural language processing (2021, neutral): “But the question is: whose standard (Eisenstein, 2013)?”
Twitter Universal Dependency Parsing for African-American and Mainstream American English (2018, background): “It contains, among other phenomena, non-standard spelling, punctuation, capitalization, and syntax, as well as Twitter-speciﬁc conventions such as hashtags, usernames, and retweet to-kens (Eisenstein, 2013).”
How Twitter is Changing the Nature of Financial News Discovery (2016, background): “Event tweet clustering can help [41, 46], but natural language processing of Twitter is substantially more difficult than news [15].”
Computational Sociolinguistics: A Survey (2016, background): “Phonology does impact social media orthography at both the word level and structural level (Eisenstein 2013a), suggesting that studies on phonological variation could inform studies based on social media text data and vice versa.”
When POS data sets don’t add up: Combatting sample bias (2014, background): “However, recent studies suggest (Eisenstein, 2013) that Twitter samples exhibit severe sample bias towards the particular time they were collected.”
Neural Unsupervised Domain Adaptation in NLP—A Survey (2020, background): “The dichotomy between canonical (typically considered well-edited English newswire) and noncanonical data arose with the increasing interest of working with social media with all its challenges related to the ‘noisiness’ of the domain (Eisenstein, 2013; Baldwin et al., 2013).”
Stylistic Variation in Social Media Part-of-Speech Tagging (2018, background): “are more authors, more kinds of authors, more varied communicative settings, fewer rules, and more stylistic variation (Baldwin et al., 2013; Eisenstein, 2013).”