Annotating Training Data

Mon, 01 Jan 0001 00:00:00 +0000

crf annotate builds CRF training data by hand. It walks a plain-text file one line at a time, shows each line in an interactive terminal, and writes the sequences you tag to an XML training file. Where Getting Started trains and runs models in code, this is the human-in-the-loop step that produces the labeled data those models learn from.

Getting Started

Mon, 01 Jan 0001 00:00:00 +0000

Conditional random fields are a type of machine learning that is used to apply a tag to tokens within a sequence when the surrounding tokens influence the tag. Conceptually, CRFs process sequences of tokens and apply tags to each token. A set of features are associated with each token, and the relationships of those features to each other are used to assign scores to each possible tag and assign the best tag to each token. Typical features include the length of the token, punctuation in the token, and tokens from the preceding and following tokens.

Documentation on CoordinateKit CRF

Annotating Training Data

Getting Started