Wednesday, December 21, 2005

What does a checker do?

As you know, I constantly witter on about this being a system and we being a team. So this is all about those strange creatures (to transcribers) who are checkers.

In an ideal world, checkers would be people who had done some transcribing. As it is not an ideal world, many checkers have not done any. Some of the most successful have just done one transcription and have probably forgotten what they learnt.

Starting up a checker is a little more complicated than getting a transcriber underway. However, if the instructions are followed, it can be done. The software is downloaded from the Free Census web site and is known as WINCC. The checker gets a zipped data file by email and loads it into the software. The task is then to go through the data, line by line, and compare it to the returns on the fiche. I cannot make the checkers look at each line and it is possible to just tab and save your way through the whole thing. I have had one or two people who appear to have done just that.

The checker should attempt to identify the transcriber’s mistakes AND correct them. This latter might seem pretty obvious – but isn’t to some people. The software enables the checker to do lots of interesting things including inserting people;complete households; or even whole pages. They can split up households and join people to households. They can alter the header data and leave lots of interesting little notes for the validater.

The checker should attempt to solve any flagged up records left by the transcriber. They can leave them for the validator and they can also flag items themselves. Of course, it is just as possible for a checker to be wrong as it is for the transcriber. However, I do not second guess checker’s decisions – well, not often. They can see the evidence and what the transcriber thought – I want them to make a decision. Flagging up the query or letting a query through is OK with me – but I would rather they sorted it out before it gets to me.

A well transcribed piece is easy to check; but as we are all human, most transcribed pieces are full of errors. The system will take care of them if we all operate it properly. Sometimes the checkers get nightmare pieces with virtually every record requiring a correction. Usually, these are repetitious mistakes and easy to correct. But each correction involves a number of key strokes or mouse clicks and it can become very boring to do ten mouse clicks for each of 5000 records!

At the end of the process the software produces another zipped file. The name has changed from censXXXX to ZZZZyyyy. This zipped file is emailed to the validater, who stuffs it into a third piece of software. You’ll have to wait for the next installment to find out about that.

1 comment:

Rick said...

You kept this very quiet - I know you had it on the stocks but hadn't realised that you had started posting.