At the 2013 BYU Conference on Family History and Genealogy, Robert Kehrer of FamilySearch presented a session titled “Effective Search Techniques and Sound Research Practices.” Kehrer is a senior product manager of search technologies for FamilySearch. Previously he worked for 12 years as a molecular geneticist identifying disease genes by building very large family pedigrees. He has an MBA and worked for 7 years at Apple managing strategic alliances and
driving market strategy in the sciences.
Kehrer said that FamilySearch has over 3.5 billion indexed names in its historical record collections. [I need to write someday about the difference between indexed records and indexed names. I think the latter can be deceiving.] FamilySearch has many unique records not found on other websites. It has particularly focused on civil and church vital records and censuses.
Order of Search Results
Search results are ordered with the best matches at the top, said Kehrer. This is determined by calculating a score for each result. Scores are calculated by comparing the information entered by the user with the information in each record. Information that matches exactly contributes to the score. Information that matches only closely contributes a little. Missing information lessens the score. Records with conflicting information are not included in search results at all.
Kehrer then explained each of these match types (exact, close, and missing).
For names, an exact match is just what you would assume. “Frank” matches only “Frank.” For exact place matches, FamilySearch uses what Kehrer called “spot of ground” match. If a location changes name, it still counts as an exact match. For example, Kingston, Addison, Vermont matches Granville. Common place name variants also count. For example, “SLC” matches “Salt Lake City.” For dates, the date in the record must be within the range specified by the user.
Name variants count as close matches for names. “Frank” matches “Frank,” “Frankie,” “Franklin,” “Francisco,” and so forth. Close place matches are spot-of-ground matches plus matches up one jurisdiction. For example, Utah County is a close match to Provo. For dates, there are no close matches.
Missing matches refer to situations in which information is specified for a field that doesn’t exist in a record. For example, a birth record doesn’t contain death date. If a user specifies death date, then the birth record might still match, but the score is decreased.
There is a special case for locations. If a user specifies a death location and a record collection doesn’t include death locations, it can match records that include the same location for birth or some other event. The location can exist explicitly in the record, or intrinsically in the context of the collection. [That’s a way of saying that records in a Michigan Death collection match Michigan whether particular records mention Michigan or not.]
There is a special case for dates. If a user specifies a range for birth and a collection doesn’t have birth dates, then other dates are examined. If a user specifies a birth year but birth years are not present in a record collection, and if marriage years are present in the collection, then the specified birth year is adjusted to an approximate marriage year and compared to marriage years in the collection.
Wildcard Characters
There are two different wildcard characters, asterisk or star (*) and question mark (?). A star matches one or more characters. For example, Stan* matches Stan, Stanley, Stanislaw, etc. *Conner matches O’Conner, Conner, etc. J*son matches Jefferson, Jamison, etc. One of Kehrer’s coworkers has learned that *lusn*k matches many common misspellings of his Polish ancestors: Czelusniak, Gelushnik, Celusnik, etc.
A question mark matches a single character. Eli?abeth matches Elizabeth and Elisabeth. Ols?n matches Olson and Olsen
Person vs. Record Search
At different times people search for different reasons. At times you search for people. “I know my grandfather. I can recognize his voice. I could pick him out of a lineup. Give me all the records you have about him.” To do this, perform a search of all collections. Specify a life event (birth, marriage, residence, death, or any) and a relative (spouse, parents, or other). This finds all matching records. Remember that missing information decreases the score, pushing results further down the list. Specifying death information will push birth certificates down the list. A parent relationship might match birth certificates and death certificates. [I might add marriage certificates and census records.]
At other times, you are looking for a specific record because you are searching for a particular record. [We genealogists are usually extending lines using proven techniques that utilize particular record types, depending on what piece of information we are looking for.]
When looking for a specific record, [say, a birth record in England,] when we specify a birth place in England, we don’t want U.S. census records [even if they match name, birth place, and birth date]. Use the “Restrict records by” feature to narrow results for a particular record from a particular place. Specify record type. Specify country and next jurisdiction (like U.S. state or English county) if known.
More next time…
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.