Monk - frequently asked questions

Subject: Login

Dear annotator,

- You can participate in the public crowd sourcing via this site, the 'Monk trainer' (http://tinyurl.com/5w5nqaf)
  It will show random words from the manuscript collections that reside on the Monk server.

- You can also log in as an 'expert trainer' with  user name/password
  on: /cgi-bin/monkweb?cmd=login

- After this, for collections that are suitable, you can perform text line transcription on:

  /cgi-bin/monkweb?cmd=editxt&db=19320&ipage=1&edit=11&ntxt=-1

  By default, Monk shows the segmented text lines in their original color version.
  For the human eye this is a bit more informative than the black-and-white version
  which Monk uses internally for word search. You can change the rendering style in the
  line transcription tool under the button [Update rendering style]

- You can also train words (i.e., label word zone images) on:

  /cgi-bin/monkweb?cmd=TrainedWords&db=19320&trainedwordmethod=Sordex&annot=all&prefix=&sortopt=sorted_hitlistsize

  This shows a hit list in which a word, letter or text pattern can be clicked and labeled individually.
  In contrast to the crowd-source labels, expert labels will be entered into the continuous learning process of Monk.
  The beneficial effects of labeling are usually noticeable after one day, on average.

This may all sound quite technical. With normal use you cannot do something that is extremely 'wrong'.
Just click around and try to label some images. A drop-down menu will show what other users
have typed as word labels, after you entered the first letter.

Best regards,
Lambert Schomaker 

Subject: crowd sourcing of medieval words

Subject: Re: leken-oogst middeleeuwse woorden

Beste meneer Schomaker,

Ik zie in de 'leken-oogst' dat afkortingen niet worden opgelost: Bij het
hedendaagse 'en' staat onder andere 'en (met een streep boven de n)'. In
het Middelnederlands is dat een afkorting voor 'ende'.

Subject: labeling

Dear Lambert,

The MONK-system is becoming more clear to me. It looks nice but I
still have a number of questions.
Do I understand it right that the abbreviated words in the labeling
are prefixed with a '@'? For example, if oviu is written
but ovium was intended by the writer, then what should I
do. We have a system for such contractions in the humanities.
For instance, 'Gijsbt' will be transcribed as 'Gijsb(er)t',
as in TEI.

Best regards,

annotator

Subject: I have labeled a word and at first it caught nice siblings in
the Sordex hit list but now no more new word instances are found. Why?

Dear user,

this can mean a number of things. Either the word is 'mined out' (there are no more instances) or, more likely, you have reached the agglomeration capacity of that method (Sordex). The trick is to go to another method (say: Qordex) and check the hit list of that same word. Even if you find a single new instance, this may help both Qordex itself, and Sordex. The principle is that of the Fahrkunst, the old mechanical up/down lift for miners. By adding a label, you push the system up, then go elsewhere and do the same. Consider the left pole in the image to be Sordex and the right pole Qordex, and you will see what I mean. The quality of the recognition is gradually pushed up:

! add new labels and shift
Sordex	Qordex

Lambert

Subject: differences between word labeling and line transcription

Dear Lambert,

I have started to browse the Cuper-Braun correspondence, that has been uploaded to MONK.
Immediately I encountered some issues:

> (1) To start with the interface: 
> To be honest, I do not understand where I need to go / where I need to click to start labeling letterforms.
> - starting with the scan, I succeed in cutting up the page in lines; 
> - every line is then accompanied by five small, square icons; 
> - the last two icons "R" and "Z" send me to a "Label Zone" (only the first one actually works on my computer/browser, since the other one requires 
> a plug-in which is missing); 

> - in the Label Zone [R], the selected line has been rendered many times, each time with another rectangular selection highlighted; the highlighted 
> selections are accompanied by suggestions for labeling;
> - if I click on one of the suggested labels, I am forwarded to a screen where the rectangular selection is magnified, and I have the option of 
> changing the label.
> BUT: most of the times, the selections are of a combination of letterforms that do not match one word, but form a selection of letters from within a 
> word, or from two words.
> Am I supposed to simply go through all the highlighted selections in a line, until I see a selection which matches exactly one word?
> I presume it is of no use to label a selection if it contains two words in a row, or a random selection of letters?

> Or is there a more convenient way of adding and correcting labels? If so, where do I click to get there?

> (2) It was easier to start editing the transcriptions of lines. 
> I wondered if the changes and additions I make to the transcriptions (and save), will be automatically used by MONK in learning, improving its 
> "mastery" of the handwriting of this collection of letters. Or do you have to order the servers at the Zernike to integrate the changes I make to the 
> transcriptions into the learning process?

> (3) One volunteer has transcribed
>
> "venusta-
> temq.," as
> "venusta-
> temp..."
> 
> The suffix "-q." looks like "-q3", and it is a very common abbreviation for "-que". Apparently your transcriber has read it as "-p[iets]"
> What is the best way to correct this? 
> Simply "temq."? (that is what I have done right now, on scan #10) 
> Or "@temque_temq3"? 
> That might make the transcription quite ponderous, seeing that such abbreviations occur so often...

> Thanks in advance,
> Yours kindly,
> ...

Subject: doelwoord met andere woorden in hetzelfde vakje

Hallo Lambert,

Nog een vraagje over het labelen in MONK:

Het komt regelmatig voor dat in de hitlist blokken staan met meer dan
een (1) woord.  Onder 'vridages' staan naast 2 groene blokken met
'vridages' bijvoorbeeld ook 3 rode blokken met 'vridages op' en 'des
vridages'.  Is het een probleem als deze woordencombinaties als één
item worden gelabeld? Zodat 'vridages op' een (1) item in de hitlist
wordt.  Of wordt binnen MONK alleen per woord gelabeld en kunnen
meerdere woorden niet als een (1) item gelabeld worden?

Hartelijke groet,  
transcribent

Subject: blue instead of green marked word zones in word labeling

Dear Lambert,

I have tried to label a number of words. For example, I changed an '11' into
'@Roman_xi_11'. After saving, the pink box with
the word became blue (Human labeled,
does not fit in current hit list) instead of green. 

Subject: General procedure around this 'Sordex'?

Dear Lambert,

 I started some word labeling but I am still in the dark what is happening in the background.
I understand that this 'Sordex' thing is very important.

Subject: Roman numbers

Dear Lambert,

I have tried to number some words. For example, I have changed a '11' (decimal
eleven) label into @Roman_xi_11 as you requested (...) but now I see blue
colors instead of the desired green color for HUMAN-labeled words.