This project adds an implementation of a fuzzy matching service to the SanteDB iCDR core instance in which it is installed. It provides:
- Services for identifying duplicate records
- Services for classifying an inbound record against the iCDR's data source
- Query Filters for fuzzy searching using the HDSI query syntax
There are two matchers that are provided in the matching project:
- DeterministicMatcher which can detect matches based on HDSI query criteria directly from the database. These matches are not scored, rather they have a probability of 0.0 or 1.0 only.
- Weighted/Scoring Matcher which uses a three stage algorithm to determine matches given a series of rules. The steps for this matcher are as follows:
- Blocking: In the blocking phase HDSI queries are run against the underlying database and are selected. This is done to reduce the number of records which need to be classified.
- Scoring: In the scoring stage, a series of transforms and measurements are done on the blocked records (from step 1), and each record is given a score. The score indicates the confidence that the record is a match.
- Classification: In the final stage, the weighted sum of all scores in step 2 are classified as: Match, Probable Match, or Non Match and returned to the caller
In SanteDB, query filters are commonly used in deterministic ways (such as =, <=, ~, etc). SanteDB 1.2 introduces the concept of extended query filters which allow consumers to indicate a transform to be performed on data prior to execution. For example, if you wanted to match a Patient's last name using Soundex instead of "LIKE" you could use:
?name.component.value=:(soundex)SMITH
Which will match SMITH, SMITHE, SMYTHE, etc.
Query filters are in the format:
property=:(extension |parameters ) operator value
| Filter | Description | Example |
|---|---|---|
:(date_diff|otherdate)timespan |
Calculates the difference between two dates | ?dateOfBirth=:(date_diff|2018-01-01)<1w Matches all patients born within one week of 2018-01-01 |
:(substr|start[,end])otherString |
Performs a partial match on a sub-portion of the string | ?identifier.value=:(substr|0,6)304-304-394 matches all patients who have an identifier starting with 304-30 |
:(levenshtein|otherString)distance |
Calculates the levenshtein difference between the property and input | ?name.component.value=:(levenshtein|Jenny)<1 matches all patients who's name is only one character different than Jenny (Jenn, Jennye but not Jennie) |
:(metaphone[|specificity])otherString |
Matches the field based on the metaphone code | ?name.component.value=:(metaphone)Smith matches any name whose metaphone code matches SMITH |
:(dmetaphone)otherString |
Matches the field based on the double-metaphone code | ?name.component.value=:(dmetaphone)Smith matches any name whose double metaphone code matches SMITH |
:(soundex)otherString |
Matches the field based on the SOUNDEX code | ?name.component.value=:(soundex)Smith matches any name whose soundex code matches SMITH |
:(soundslike|otherString) |
Matches the field based on the currently configured phonetic algorithm handler (lets server decide the algorithm) | |
:(phonetic_diff|otherString[,algorithm])distance |
Matches a field based on phonetic difference to another code using metaphone (default) or using soundex or dmetaphone | ?name.component.value=:(phonetic_diff|SMITH)<2 Matches any name where the metaphone code is only 1 character different |
:(alias|otherString)relevance |
Matches records which are an alias | name.component.value=:(alias|Will)>0 will match Will, William, and Bill |
Up to date documentation can be found on the SanteDB Wiki