Sourceform

Sourceform is a term used in linguistics and data processing to refer to the original input representation of a word or data item before any normalization, analysis, or transformation is applied. In natural language processing, sourceform is often treated as the input string and may be used interchangeably with terms like surface form or input form, depending on the source.

In linguistic analysis, the sourceform is the form of a word as it appears in the text, including inflectional endings, capitalization, and punctuation. It is distinct from the lemma, the canonical dictionary form used for indexing and morphological analysis. Tools that perform lemmatization or stemming typically map the sourceform to a lemma or stem. However, terminology varies: some authors use sourceform and surface form interchangeably, while others reserve one term for the visible text and the other for internal representations used by a tool.

In data pipelines, sourceform denotes the raw data state before transformation. For example, a source-form CSV or JSON is the original data file that is then cleaned, normalized, and imported into a target schema. The term emphasizes the starting point of processing, as opposed to the transformed or normalized form.

Origin and usage: The phrase does not have a single standard definition across all disciplines, and its precise meaning depends on context. It is most common in NLP and information extraction contexts but is also used in data integration and digital humanities.

See also: surface form, lemma, stem, lemmatization, morphological analysis, data normalization.