Search for content in message boards

Announcing upcoming test for search relevancy

Announcing upcoming test for search relevancy

Posted: 26 Oct 2012 4:02PM GMT
Classification: Query
Hi all,

We are in the process of rolling out our next test to verify the default behaviors on new search are the correct default behaviors to help you find the records for your ancestors most easily. You may recall that our earlier test (http://boards.ancestry.com/topics.ancestry.ancimprovements/1...) indicated we should remove Soundex matches from the default experience. This test will help us determine if we should, by default, tighten search results even more.

The changes we will be testing include:

1) Whether Phonetic algorithm matches should be included by default on new search. For this test, we will be comparing how successful searchers are under the current default behavior (with Phonetic matches on) vs. Phonetic matches off (we'll still include "similar" matches).

2) Whether there should be some degree of first name matching. Currently the website does not require any matching on first name at all. This is why you can get results for Peter Blow when you search for Joe Blow. For this test we will compare successes from the current default behavior to one where the results must match exactly, or phonetically, or similar, or because of initials.

We will be analyzing performance of the different experiences across different user types and deciding upon the configuration that works best for the most users.

As with our last test, this test reflects our commitment to taking a reasoned approach to significant changes to the user experience.

Thanks,

John
Ancestry.com
Search Product Team

Re: Announcing upcoming test for search relevancy

Posted: 26 Oct 2012 7:16PM GMT
Classification: Query
Thanks, John.

Do you have a scheduled start date for the test? How long will it run?

Thanks.

Re: Announcing upcoming test for search relevancy

Posted: 26 Oct 2012 10:05PM GMT
Classification: Query
Oh, dear, John, this is asking for trouble (for my searching as well as otherwise). How do you plan to phonetically match these:

Aung San Suu Kyi
Wojtowicz
Kune when pronounced "Carney"
Rzeppa
Qianlong (in Pinyin system)
--not to mention many Welsh and Cornish names

Would much rather see:

1) Eliminate high ranking for result that shares only middle initial with target name: when I search for Daniel E. XXXXX, I do not want all the Mary E., Joseph E., Hieronymus E., John E., etc. results to precede the ones for Daniel E.

2) Date bracketing that ~works~ when search from tree has birth/death dates, so for person born 1869 I do not get ANY Civil War, Revolutionary War, War of 1812, 1850 Census, etc. results. And for person died 1869 there should be **no** 1880, 1900, 1910, 1920, 1930, 1940 US Census results, nor any for World War I and World War II draft registrations.

3) When restricting to place (from drop-down) Licking County, Ohio, USA as publication place in newspaper search I get lots of results for Lima, Ohio. There are towns named Lima in four other Ohio Counties, but none in Licking County.

Re: Announcing upcoming test for search relevancy

Posted: 26 Oct 2012 10:30PM GMT
Classification: Query
Edited: 26 Oct 2012 10:31PM GMT
I hope the first name part of this test goes well. I think I would love to see, books, newspapers, land records, etc have a good search including first names.

Re: Announcing upcoming test for search relevancy

Posted: 26 Oct 2012 10:38PM GMT
Classification: Query
Maybe I'm misunderstanding, but phonetic matching is already included in search results. We're not introducing any new variations to search results. (You are correct that phonetic, like Soundex, is better for Western European names than other types).

As for the things you would rather see. We are taking a methodical approach and will do further tests as we go along.

I do want to clarify one thing you mention. If you enter a death date of 1869, you should never see a census for 1880+. Our lifespan filtering will filter that out (you may see 1870, however). I think the issue you are talking about is when a record doesn't have an explicit life event date associated with it. Our lifespan filtering uses life events (birth, marriage, residence, etc.) to determine whether a record could have been associated with a person. If a record doesn't have an event date, It doesn't get excluded. The problem is that some of these records have an implied date that clearly (to the user) makes the result irrelevant. The problem you mention is valid, and is somethign we'd like to address, it just won't occur (unless a bug) with the specific databases you mentioned.

John

Re: Announcing upcoming test for search relevancy

Posted: 26 Oct 2012 10:52PM GMT
Classification: Query
John,

Sorry, but I have to disagree - the lifespan filtering almost never works. In fact, I had asked if it was ever implemented, since I see so many "wrong" returns. When I asked Tech Support, they said they did not know what I was talking about.

When I search in military records for someone born in 1895, if lifespan filtering is working, I should not see Civil War service records, but I do.

Re: Announcing upcoming test for search relevancy

Posted: 27 Oct 2012 8:04AM GMT
Classification: Query
John, you said "If a record doesn't have an event date, It doesn't get excluded."

There are plenty of items having dates that are not excluded by comparing with vital dates of specific tree person.

In a huge number of instances the record dates simply have not been indexed, however, so the search engine wonkily lists them as results. Say, all of the WWI and WWII draft registration items, all Pension Application File Indexes . . . and many others that could readily be excluded as a group from results for a person born at a certain time or died at a certain time.

Re: Announcing upcoming test for search relevancy

Posted: 27 Oct 2012 2:16PM GMT
Classification: Query
The search relevancy will always be crap until Ancestry fixes the indexing in numerous databases. The system can't return what it can't find.

How many times do I have to post about the Drouin Collection's indexing? Months have gone by and the indexing remains so convoluted and erroneous is can't be searched effectively, no matter what relevancy tinkering is done.



Re: Announcing upcoming test for search relevancy

Posted: 31 Oct 2012 3:50PM GMT
Classification: Query
"If you enter a death date of 1869, you should never see a census for 1880+."

I just did a search regarding a person who died in 1933. The top three results were for 1940 US Census, wrong places, wrong names except surname.

I am sometimes getting the same sort of mismatches as 'hints'.

Re: Announcing upcoming test for search relevancy

Posted: 31 Oct 2012 4:14PM GMT
Classification: Query
"If you enter a death date of 1869, you should never see a census for 1880+."

BUT, I should see the probate files which always take place after death, sometimes many years later.
per page

Find a board about a specific topic