A case-restoring spell corrector, also called a truecasing corrector, only restores the case and does not change anything else, that is, it does not correct spelling errors. This is very useful when dealing with low-quality text from transcriptions, automatic speech-recognition output, chat logs, and so on, which contain a variety of case challenges. We typically want to enhance this text to build better rule-based or machine-learning systems. For example, news and video transcriptions (such as closed captions) typically have errors, and this makes it harder to use this data to train NER. Case restoration can be used as a normalization tool across different data sources to ensure that all the data is consistent.