NOBLE – Flexible concept recognition for large-scale biomedical natural language processing

BMC Bioinformatics

Table 4 Examples of NOBLE matching strategies produced by combinations of matching options

Use Cases		Combination of matching options
Task	Description	Subsumption	Overlap	Contiguity	Order	Partial
Best match	Provides the narrowest meaningful match with the fewest candidates. Best for concept coding and information extraction.	Yes	Yes	Yes (gap = 1)	No	No
All match	Provides as many matched candidates as possible. Best for information retrieval and text mining.	No	Yes	No	No	No
Precise match	Attempts to minimize the number of false positives by filtering out candidates that do not appear in exactly the same form as in controlled terminology. Similar to best match, but increases precision at the expense of recall.	Yes	No	Yes (gap = 0)	Yes	No
Sloppy match	Allows matching of concepts even if the entire term representing it is not mentioned in input text. Best for concept coding with small, controlled terminologies and poorly developed synonymy.	No	Yes	No	No	Yes

For contiguity, the gap indicates the number of words (not counting stop words) that can occur in-between words that make up a valid term

ISSN: 1471-2105