📄 Jaccard Index Words¶

JaccardIndex metric that operates on prediction and references that are strings.

It first splits the the string into words using space as a separator.

For each prediction, it calculates the ratio Intersect(prediction_words,reference_words)/Union(prediction_words,reference_words). If multiple references exist, it takes the best ratio achieved by one of the references.

metrics.jaccard_index_words

JaccardIndexString(
    splitter=RegexSplit(
        by="\s+",
    ),
)

[source]

from unitxt.string_operators import RegexSplit

Explanation about JaccardIndexString¶

Calculates JaccardIndex on strings.

Requires setting the ‘splitter’ to a FieldOperator (such as Split or RegexSplit) to tokenize the predictions and references into lists of strings tokens.

These tokens are passed to the JaccardIndex as lists.