In Philter, a span is a piece of the input text that Philter has identified as sensitive information. A span has a start and end positions, a confidence, a type, and other attributes. Ideally, each piece of identified sensitive information will only have a single span associated with it. In this case, the type of sensitive information is unambiguous. The goal of span disambiguation is provide more accurate filtering by removing the potential ambiguities in the types of sensitive information for duplicate spans.