Pagina iniziale› Gruppi› Full Text Search Feedback

Full Text Search Feedback

Partecipa

Record sets removed from

marleenvanhorne2
marleenvanhorne2 ✭
6 December in Social Groups
0

Commenti

  • marleenvanhorne2
    marleenvanhorne2 ✭
    6 December

    When record set moved to Full Text, it was removed from the catalogue, and was not easily usable, because full text does not use fuzzy searches and is a stupid search.
    The item in question is Genealogical Gleanings from Book No. 2* of Conveyances, Brooklyn, Kings Co., N.Y. At least with an index in the record set you could read the names and find the record, no matter how the name was spelled. By not using fuzzy search, many records will be unfindable.
    Can't give a URL, because the record set is no longer in the record set for Kings Co. Colonial New York British America.

    0
  • David A Wilson
    David A Wilson ✭✭
    6 December

    @marleenvanhorne2 Can you give me an example of what you consider a "fuzzy" search?

    The instructions list these options. Do they not meet your definition of a fuzzy search?

    Tips for successful full-text searching

    Too many results? Try these strategies to focus your search:

    Find an exact word or phrase

    Use quotation marks. Example: "Henry Jones"

    Include a specific word or phrase

    Use the + symbol. Example: +Judith

    Exclude a specific word or phrase

    Use the - symbol. Example: -John

    Find different spellings

    Use the ? symbol. Example: Jens?n will return Jensen and Jenson

    Find different forms of a root word

    Use the * symbol. Example: Car* will match car, cars, carriage, carpenter, etc.

    0
  • marleenvanhorne2
    marleenvanhorne2 ✭
    6 December

    ON Ancestry [hate to give A credit for anything] for a given name you can specify "Exact, sounds like, similar & initials" and for a surname, "Exact, sounds like, similar & soundex."

    0
  • Mark McKenzie_1
    Mark McKenzie_1 ✭✭✭
    7 December modificato 7 December

    @marleenvanhorne2 Full Text Search requires you to do the 'fuzzy' thinking yourself and doing it well is important for improving your chances for success. The search 'options' @David A Wilson mentioned are useful but only go so far and are not that 'flexible'.

    Say you are looking for a Steve Johnson and that is how his name is presented in sources you have already found—- In FTS if you enter, Johns?n, as one of your keywords, it will return image transcripts that have the word transcribed as Johnson, Johnsan, Johnsen, Johnsin etc in addition to Johnson.

    Although the wildcards can be helpful, FTS has one current drawback, you cannot use the ? in a phrase, e.g. +"Steve Johns?n". You must use a keyword search such as this —- +Steve +Johns?n which will give less than useful records. Any image transcription result would have those two words anywhere in the transcript and quite likely completely unrelated to each other.

    Same limitation applies when using the asterisk. Using Johns*n as a keyword will generate transcript results with — Johnson, Johnston, Johnsten, Johnstin, etc. But unfortunately you can't use this wildcard in a phrase +"Steve Johns*n"

    This limitation has been highlighted, but it may be in the 'too hard to do' file from a software engineering perspective. I hope not as it would be a great help!

    So for now, you will need to come up with your own "Exact, sounds like, similar & soundex." and then systematically do multiple searches e.g. +"Steve Johnson", +"Steve Johnsen", +"Steve Johnston"

    And as they say…. 'But wait there's more'….

    You need to do some fuzzy thinking about how his name may appear in the records and again do individual searches e.g. +"Steve Johnson", +"Steven Johnson", +"Stephen Johnson", +"S Johnson", +"Johnson Steve", +"Johnson S". That was just using the surname, Johnson, so again more searches needed to cover the surname variations. Getting to be a long list of keyword search terms using various combinations of both given and surnames

    It would be nice if there was an 'AI Assistant' when using FTS, which would generate all the plausible given and surname keyword combinations for you and maybe even give you an option to 'launch' multiple FTS using the provided variations. Probably asking a bit much here.

    Thought I'd add a fuller explanation about FTS, not that you need it obviously, but hopefully others could use the info

    0
  • marleenvanhorne2
    marleenvanhorne2 ✭
    7 December

    NO! AI needs to do the fuzzy thinking, that is the whole point of having it. A human being should not have to think of the multitude ways names could be misspelled, in order to get a match. Really fuzzy searching is the whole point of Artificial Intelligence. If it cannot do that, then it is really STUPID.
    I have 41 years experience researching in these documents and have found records when I didn't even know the name of person I was looking for. The point, however, is that most people are not that gifted or persistent. If you leave people with the impression that AI has found all the records when it is doing a stupid search, you are doing a great disservice to the genealogical community.

    0
  • Mark McKenzie_1
    Mark McKenzie_1 ✭✭✭
    7 December

    To say that Full Text Search is AI is conferring upon it more capability than it really has…. It has pretty simple 'skills' and limited to simply looking thru transcribed text of its processed images for your keyword[s]. That's the extent of it. Hence your disappointment. But really back to your original post

    '…. When record set moved to Full Text, it was removed from the catalogue… the record set is no longer in the record set for Kings Co. Colonial New York British America.

    This I do wonder about as I don't think that Full Text Search processing should have caused that to happen… Maybe this should be posted in the general Community area instead of the Full Text Search Feedback Group. Going that route might help you figure out where those records may have gotten to….

    0
Clear
No Groups Found

Categorie

  • Tutte le Categorie