tRuEcasIng

By L. Lita, Abraham Ittycheriah, Salim Roukos, N. Kambhatla. 2003

138 citations recorded by the source corpus.

Truecasing is the process of restoring case information to badly-cased or non-cased text. This paper explores truecas-ing issues and proposes a statistical, language modeling based truecaser which achieves an accuracy of (cid:24) 98 % on news articles. Task based evaluation shows a 26% F-measure improvement in named entity recognition when using truecasing. In the context of automatic content extraction, mention detection on automatic speech recognition text is also improved by a factor of 8. Truecasing also enhances machine translation output legibility and yields a BLEU score improvement of 80 : 2% . This paper argues for the use of truecasing as a valuable component in text processing applications.

Subfields: Artificial Intelligence, Information Systems, Natural Language Processing

How researchers see this paper

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

In Artificial Intelligence

1. Philipp Koehn
2. Luke S. Zettlemoyer
3. Christopher D. Manning
4. Jianfeng Gao
5. Mohit Bansal
6. Graham Neubig
7. Noah A. Smith
8. Yejin Choi

In Information Systems

1. Christopher D. Manning
2. Matt Post
3. Danqi Chen
4. Hieu D. Hoang
5. Jörg Tiedemann
6. Jaewoo Kang
7. Dan Roth
8. Kenneth Heafield

In Natural Language Processing

1. Philipp Koehn
2. Luke S. Zettlemoyer
3. Graham Neubig
4. Christopher D. Manning
5. Mohit Bansal
6. Rico Sennrich
7. Kevin Duh
8. Yonatan Belinkov

Citations in the corpus

• Learning Dense Representations of Phrases at Scale (2020, positive): “Since SpanBERT only supports cased models, we also truecase the questions (Lita et al., 2003) that are originally provided in the lowercase (Natural Questions and WebQuestions).”
• Robust Named Entity Recognition with Truecasing Pretraining (2019, methodology): “Further innovations include lattice-based decoding with language models (Lita et al. 2003), and truecasing as sequence tagging (Chelba and Acero 2006), with applications to machine translation (Wang, Knight, and Marcu 2006) and social media (Nebhi, Bontcheva, and Gorrell 2015).”
• Factored Translation Models (2007, methodology): “To demonstrate the versatility of the factored translation model approach, consider the task of recas-ing (Lita et al., 2003; Wang et al., 2006).”
• The JHU Machine Translation Systems for WMT 2016 (2016, methodology): “casing (Lita et al., 2003), we use the following heuristic for these languages.”
• Phrasal: A Toolkit for New Directions in Statistical Machine Translation (2014, methodology): “The LM-based truecaser (Lita et al., 2003) requires an LM estimated from cased, tokenized text.”
• Edinburgh Research Explorer Stanford University's Submissions to the WMT 2014 Translation Task (2014, methodology): “The English output was recased using a language model based recaser (Lita et al., 2003).”
• Are Multilingual Neural Machine Translation Models Better at Capturing Linguistic Features? (2020, methodology): “To preprocess our data, we first run a truecaser (Lita et al., 2003) before splitting into subword units using BPE (Lita et al., 2003).”
• Temporal Effects on Pre-trained Models for Language Processing Tasks (2021, methodology): “Truecasing with New York Times Truecasing (Gale et al., 1995; Lita et al., 2003) is the task of case restoration in text.”