Language Normalisation Of Noisy Text Data
This paper addresses the issue of language normalization, an important problem in natural language processing.
Facebook, Twitter provides access to large volumes of data in real time, but is notoriously noisy, hampering its utility for
NLP. By language normalization, we mean converting ‘informally inputted’ text into the structured form, by removing
‘noises’ in the text. It includes detection of ill-formed words, detecting paragraph and sentence boundaries in the text.
Previously, text normalization issues were often undertaken in an ad-hoc fashion or studied separately. This paper first gives
a formalization of the entire problem. It then proposes a Knowledge based approach to perform to make the text data errorfree.