02 April 2014

Does Unexact at FamilySearch Include Soundex?

Many genealogists leave "exact" boxes unchecked when performing searches because more is better. Many genealogists also have unspoken assumptions about what searches do when "exact" boxes are checked. Those assumptions can get one in trouble.

Apparently an "unexact" search at FamilySearch does not include Soundex variants--at least not always.

The last names of Goldenstein and Goldenstien are Soundex equivalent (G435-check it out here). There's two reasons for this:
  • the only differences in the two spellings involve vowels and Soundex ignores vowel sounds
  • the differences in the names are after the third consonant sound (not counting the initial letter) and Soundex does not consider any consonant sound after the third non-initial letter. Even if consonants this "far down" in the name are switched, the Soundex code does not change. 
Consequently one would expect  a search for "Voke Goldenstein" with exact unchecked at FamilySearch in their 1880 index to find an entry for Voke Goldenstien. After all, those last names are Soundex equivalent and an "unexact" search "should" catch Soundex equivalents at FamilySearch.

Nope..not quite.

It does not as shown below. The search for Voke Goldenstein (with last name unexact) finds no results:


A search for "Voke Goldenstien" does find the entry.

The fact that the first name is checked as exact does not matter.

It is pretty clear that FamilySearch does not view the last names of Goldenstein and Goldenstien as "similar" enough to return the same results (with "unexact" unchecked) for these searches.

The search for William (checked) and Goldenstein (unchecked) returns 29 hits as shown below:


The search for William (checked) and Goldenstien (unchecked) returns 29 hits as shown below:


Clearly FamilySearch does not consider these names "close enough" to bring all of them up when an "unexact" search is the desired search option.

One would not expect the transposition of two vowels towards the end of a name of this length to matter, especially when the vowels in question are "i" and "e."

It's always advised to experiment with any "unexact" search. If you didn't code it, you don't know how it searches.

Post a Comment