Ancestry Magazine

The new search engine at Ancestry.com will rank your results and place the best possible matches at the top. You’re gonna love it.

Ancestry.com recently unveiled a new “best matches” relevance-ranked search method to add to its traditional “exact matches” search. If genealogy is like finding a needle in a haystack, you can think of this new ranked search method as a way to dig through less hay and find more of those precious needles.

This new ranked search method helps you find your ancestors quickly and easily because it brings up the best possible matches first in your list of search results. In addition, it gives you the option of performing event-based searches, meaning that you can specify a date and place of birth and/or a date and place of death in your search criteria. And, to make it even easier for you to decide where to look first, the search results page reveals more information about your matches right away. You can now see the most relevant matches from all Ancestry.com databases on the initial search results page, complete with basic information about birth, marriage, and death events.

How Does It Work?
Relevance is determined by comparing your search criteria with the information found on a genealogical record. The closer the match, the more relevant that record is to your search. Matches can be “close” in a number of different ways on genealogical records. For example, the names Johnson and Johansen are phonetically c lose to one another. Similarly, the names Bob and Robert are closely related as cultural variations of each other. New Hampshire is closer geographically to Maine than to California. Likewise, 1907 is closer to 1908 than to 1903 in a chronological sense.

The ranked search automatically returns alternate spellings and abbreviations for your ancestor’s name(s). An exact name match is the closest match, and therefore the most relevant, followed by name matches based on typographical, cultural, and phonetic closeness. Additionally, this search method uses chronological closeness as a factor in determining the relevance of date matches.

Example 1: James Otho Sharbrough
A ranked search for James Otho Sharbrough (born in Mississippi in 1882, died in Texas in 1936) returns three appropriate records in the top seven results. The first appropriate match (and the first on the list of possible matches) is a World War I Draft Registration record for “James Otho Sharbrough.” The second appropriate match is a 1920 census record for “J O Sharbrough.” The third appropriate match is a 1930 census record for “James O Sharbrough.”

Notice that the relevance ranked search automatically searched for different potential abbreviations of the name James Otho Sharbrough. This means that you don’t have to run three different variations of exact match searching to find these three records.

Also, notice that the birth date on James O Sharbrough (the seventh on the list of possible matches) is estimated to be 1886, whereas the search criteria was for an 1882 birth. Although the record appears to be “off” by four years, based on previous research, this is the correct James. Either James forgot his birth date or the census taker made a mistake on the census form. Because the relevance-ranked search looks for close matches in terms of chronology, the 1886 record is a good match.

You’ll notice that several of the other records returned in the first seven matches of this example were for a James Scarbrough, a name that is very similar to James Sharbrough. These matches were returned because the relevance-ranked search engine automatically looks for alternate or similar name spellings.

Example 2: Charles Peter Splonskowski
A ranked search for Charles Peter Splonskowski (born in South Dakota in 1889, died in South Dakota in 1969) returns eight appropriate records in the top ten results, the other two were actually records for Charles’s spouse, Marcella. Notice that not only did this search return appropriate name abbreviations, it also returned several records where Charles’s name did not exactly match the search criteria (due to transcription errors or inaccuracy on the records), but the matches were still appropriate.

For example, a record for Charles Paul (not Peter) Splonskowski was returned from the Family Data Collection. This is the right Charles, but the source record incorrectly lists his middle name as Paul rather than Peter.

Another appropriate match is a 1920 census record for Olar P. Splonskowski. In this case, the original census record appears to read “Olar” rather than Charles. Based on other information already known about his household, this record really is for Charles.

Finally, the appropriate match from the 1930 census is for Charles Splonskoreosk. In this case, the name on the census record was incorrectly indexed on the Ancestry.com website. The relevance-ranked search was able to return misspellings and other close name matches where the other search criteria (birth date/place and death date/place) were strong matches. The records for Charles’s wife, Marcella Devine, were returned because Charles was mentioned somewhere on the record.

Example 3: George Feulner
A ranked search for George Feulne r (born in Germany in 1856, died in Utah in 1931) yielded four appropriate records in the top five results. One of the appropriate records that was returned was for Georg (not George) Feulner, the name George originally went by in his homeland of Germany. The relevance-ranked search automatically searched for common nicknames and language variants of first names in order to produce this result. This means you don’t have to think of every possible nickname or foreign spelling of a name in order to achieve success. In the future, Ancestry.com plans to expand this foreign spelling concept to include geographical place names. For example, a search for Germany could return place names like Deutschland.

Don’t Forget Exact-Match Searching
To get the best possible results on a ranked search, type in as much information as possible. The more search criteria the search engine has to match against, the more likely it is to pull the most appropriate records up to the top of your results list. Not sure about an exact date? Take an educated guess. As long as you’re within a few years, you’ll get much better results than if you leave a date field blank.

If you’ve filled out all of the information you possibly can on a ranked search and you still can’t find the appropriate record(s), consider giving exact-match searching a try. Remember, exact-match searching functions much differently than relevance ranked searching, and thus requires a vastly different search strategy for success.

In exact-match searching, each match must exactly match all of the search terms you enter. You should probably start an exact match search with only one or two broad search criteria (e.g., a surname and a location). If you get too many results, gradually add more criteria to narrow your search. If you get too few results, drop one or more of your search criteria to broaden your search.

Although exact-match searching d oes not automatically search for abbreviations and alternate spellings like relevance-ranked searching, you have the option of using the Soundex feature, which returns many phonetically similar surnames. Use the Soundex feature if you wish to broaden your results. Additionally, exact-match searching supports the use of wildcards (special symbols such as the asterisk (*) and the question mark (?) that can be used to represent an unknown letter or group of letters in a word). You can use wildcards in an exact-match search to manually hunt for alternate name spellings.

What’s Next?
Plans for subsequent versions of ranked search include adding more events to the advanced search form, allowing users to specify gender and maiden name to better find records for women, and adding geographical closeness as a factor in determining the relevance of location matches. All of these upgrades, coupled with the continued dedication of Ancestry.com to improve the other matching capabilities of the search engine, will ensure that the new ranked-search method will only get better with time. Good-bye hay, hello needles!

Kendall Hulet is a product manager for MyFamily.com, overseeing the Search user experience. He has written several articles for academic journals that relate to Internet business practices.