spaCy/website/api/_annotation/_biluo.jade

//- 💫 DOCS > API > ANNOTATION > BILUO

+table(["Tag", "Description"])
    +row
        +cell #[code #[span.u-color-theme B] EGIN]
        +cell The first token of a multi-token entity.

    +row
        +cell #[code #[span.u-color-theme I] N]
        +cell An inner token of a multi-token entity.

    +row
        +cell #[code #[span.u-color-theme L] AST]
        +cell The final token of a multi-token entity.

    +row
        +cell #[code #[span.u-color-theme U] NIT]
        +cell A single-token entity.

    +row
        +cell #[code #[span.u-color-theme O] UT]
        +cell A non-entity token.

+aside("Why BILUO, not IOB?")
    |  There are several coding schemes for encoding entity annotations as
    |  token tags.  These coding schemes are equally expressive, but not
    |  necessarily equally learnable.
    |  #[+a("http://www.aclweb.org/anthology/W09-1119") Ratinov and Roth]
    |  showed that the minimal #[strong Begin], #[strong In], #[strong Out]
    |  scheme was more difficult to learn than the #[strong BILUO] scheme that
    |  we use, which explicitly marks boundary tokens.

p
    |  spaCy translates the character offsets into this scheme, in order to
    |  decide the cost of each action given the current state of the entity
    |  recogniser. The costs are then used to calculate the gradient of the
    |  loss, to train the model. The exact algorithm is a pastiche of
    |  well-known methods, and is not currently described in any single
    |  publication. The model is a greedy transition-based parser guided by a
    |  linear model whose weights are learned using the averaged perceptron
    |  loss, via the #[+a("http://www.aclweb.org/anthology/C12-1059") dynamic oracle]
    |  imitation learning strategy. The transition system is equivalent to the
    |  BILOU tagging scheme.