drjwbaker/Thursday (DHOxSS 2013)

## Thursday (DHOxSS 2013)
[live notes, so excuse the errors, omissions and personal perspective]
***This work is licensed under a Creative Commons Attribution 3.0 Unported License.***

[live blog] #DHOxSS, Christine Madsen and Matthew McGrattan, Digital Library Technologies and Best Practice
Christine Madsen, Part 1: Deconstructing Digital Libraries (or: why you should work with your library!)
Bodleian Digital Library Systems & Services

Talk about what we are doing but also libraries, and how work in libraries slightly different to the academic perspective on digital projects.
Bodleian Digital Library Systems and Sercices provide electronic core to all Oxford libraries.
15 years of experience digitising collections; 15-20 ongoing projects at present.
But not all traditional digitisation: Early Modern Letters Online a flagship project around metadata.
26k comments: real success thre.

"Our work in the Digital Humanities has changed the way I think about libraries and what they do"

So what do libraries do?
We describe salient features of content.
> what has changed is a move to toward that content as data.
We store artefacts of intellectual discourse.
Libraries support knowledge creation.
> back when we built libraries that looked like the Bod we knew how to do this. Position libraries as central to learning.
>> eg when Thomas Bodley gave his collection of books (and some money) he was heralded as having saved learning.
> Now... 0100101010101 ... sorting information into buckets
>> We are here because of Father Busa, but while it took him 56 years we expect an instant output. Slow research no longer acceptable.
Libraries are open.
> but a push to even greater opennnes right now, and openness is tied in with collaboration. And collaborative standards
Libraries produce outputs.
> and what they are is changing.
Libraries evolve.
> Libraries do not have a great history of making rapid change - like to think long and hard - but speed now important
Libraries provide longevity

How to stay relevant?
> question of how to maintain our role as custodians in a changing environment
>> working with academics, getting involved in projects as early as possible, be open on issues, visible on cost.
> hosting vs archiving
> pay attention to users as well as project requirements.

Seperate content from interface.
Content is to be saved, interface is not built forever. Interfaces ephemeral.
Content sits on top of metadata and infrastructure.
[I think this is really important]

Matthew McGrattan, Part 2: Building a digital collection.
Digitisation Service Manager & Imaging Specialist

The Digital Turn... access, opporunities, dissemination
The Basics of starting a digital project... What do you have? Who is going to use it? How are you going to build it?

What
Content, plus what is going to sit on, around and below the content

Who
Specialists? Technical terminology may be off-putting depending on target audience.
But of course also non-specialists > striking the right balance.
Machines. Machine readability important.

How
Driven by what and who.
Example of questions can be seen when we think about digitising images.
- What format - lossy or lossless?
- OCR needs 300pi or above.
- What is preservable? Future-proofing and format integrity.
- Metadata: stored within the image or alongside?
- Display: flat, panning/zooming, comparative interface, image manipulation.

A warning about standards!
Standards proliferation doesn't usually give you anything extra, just more standards
And yet we do want to avoid silos, so unification of standards is worth the effort - Digital.Bodleian project.
Used a stack of services from infrastructure to user interface: combintion of standard services and developed tools.

QA
Departments territorial, libraries neutral: how does this work at the Bod?
We can't tell anyone to do anything, so we create tools/environment for collaboration.
We work on lots of these projects, individual scholars may be working on one for the first time.


Javier Pereda, Interactivity and Inferfaces

Way of trying to make us think differently about our data.
Interaction not just between people, but also machines/systems. Reciprocal interaction between entities.
Interfaces are not just computers...
So think about interfaces outside of the context of the computer to create good digital interfaces.
QWERTY Keyboard an example of an analogue interfact transposed on the digital when the analogue complexities that dictated the original design have been solved.
Command line the origin of how we talked to computers.
This evolved into the text menu interface: but now the BIOS is a cumbersome interface.
Graphical User Interface. GUIs give simple or elaborate visual representations of the object.
Studies in human computer interaction take this further...
Natural User Interface. Natural human interfaces.
- Multi-sensory interfaces (eg iPhone gestures).
- Human like interaction. Make the actions similar to what a human would do.
- Biometric interaction.
- Invisible computing.
Tangible interfaces
> require less learning than a GUI.
> user just thinks for his or herself, doesn't need to know how the interface works.

The Internet of Things
Allows objects to react, with or without us.
Though codependency important: if we didn't use wikipedia, wikipedia wouldn't interact with other objects/things.
Stop thinking of computer as the screen which presents information, rather think how best to present information.

Need to think beyond the keyboard and the mouse and the gestures on a tablet.
And interfaces which people aren't afraid to use, unlike the keyboard/mouse.

Users learning from mistakes... do they want to? what kind of mistakes are acceptable/invisible?
[this seems a bit like making every interface as fuzzy as possible given the technology available at a given moment]


Ségolène M. Tarte, Working with Digital Images

Why images? Conserve, preserve, cuarate historic documents.
But also to analyse, study and interpret them & document, present and disseminate.

What are digital images?
x, y + 0-255 (for greyscale)
alpha is a transparency value.
HSL: hue, saturation and lightness on 0-1 scale > if l=0 colour is black, if l=1 colour is white, if s=0 colour is grayscale.
Mach band effect [now that is fascinating] > shows that perception of brightness within greyscale helps us detech changes.
So, when processing/digitising an image we are making an act of interpretation. Because seeing is subjective and image capture interpretive.

Histogram based processing of images: counting values (say between 0 and 255 on a histogram)
(Logarithmic count scale more useful at times than linear count scale)
When we adjust the brightness, so we can see better, the histogram changes.
- we take the whole thing and push it along the scale.
When we adjust the constrast we strech out the grey values, redistribute them along the scale.
(though usually you adjust a bit of both)

Image segmentation

Problem: you know what you are looking for but you need to really know what you are looking for to be able to find it...
How then?
Identify features within an image: such as blobs, lines, edges (things that are tripped up on along the way)
> related to brightness persception, so greyscale images usually used for detection.
Or region identification: so for example, an area not covered in ink (sweeping up similar things)
Sobel filters help scratch detection.
Thresholding: between that and that make it white, between that and that make it black.

Processing images is modifying them, interpreting them (so keep the original!)
Understanding the (ofte black-boxed) image processing options helps justify choice and make expectations explicit
> the act of interpretation then becomes more shareable, reproducible, and teachable.

GIMP
Windows > Dockable dialogs > Histogram
Colours > Hue > move saturation to -100 to make things greyscale.
Colours > Desaturate
Colours > Brightness-Contrast
... histogram here is useful for keeping track on precisely what we are doing


RTI Viewer

Uses multi image capturing processed using Polynomial Texture Mapping (PTM) to display the image.
Move light around the object to simulate depth, and simulate our looking around an object to understand it.
Coefficient means that only the equilavent of 9 greyscale images are saved (RGB (so 3) plus 6 coefficients)

Multi-image use in GIMP
Blend modes... layered stack. Use Windows > Dockable Dialogs > Layers and then Mode to combine/compare images.
Chances are that blending them (and lightening from a white background) will illuminate the text.
Do lighten only with black bottom layer, and darken only with white bottom layer, then stack two compiled images using difference.


***This work is licensed under a Creative Commons Attribution 3.0 Unported License.***

<a rel="license" href="http://creativecommons.org/licenses/by/3.0/"><img alt="Creative Commons License" style="border-width:0" src="http://i.creativecommons.org/l/by/3.0/88x31.png" /></a><br />This work is licensed under a <a rel="license" href="http://creativecommons.org/licenses/by/3.0/">Creative Commons Attribution 3.0 Unported License</a>.
	[live notes, so excuse the errors, omissions and personal perspective]
	*This work is licensed under a Creative Commons Attribution 3.0 Unported License.*

	[live blog] #DHOxSS, Christine Madsen and Matthew McGrattan, Digital Library Technologies and Best Practice
	Christine Madsen, Part 1: Deconstructing Digital Libraries (or: why you should work with your library!)
	Bodleian Digital Library Systems & Services

	Talk about what we are doing but also libraries, and how work in libraries slightly different to the academic perspective on digital projects.
	Bodleian Digital Library Systems and Sercices provide electronic core to all Oxford libraries.
	15 years of experience digitising collections; 15-20 ongoing projects at present.
	But not all traditional digitisation: Early Modern Letters Online a flagship project around metadata.
	26k comments: real success thre.

	"Our work in the Digital Humanities has changed the way I think about libraries and what they do"

	So what do libraries do?
	We describe salient features of content.
	> what has changed is a move to toward that content as data.
	We store artefacts of intellectual discourse.
	Libraries support knowledge creation.
	> back when we built libraries that looked like the Bod we knew how to do this. Position libraries as central to learning.
	>> eg when Thomas Bodley gave his collection of books (and some money) he was heralded as having saved learning.
	> Now... 0100101010101 ... sorting information into buckets
	>> We are here because of Father Busa, but while it took him 56 years we expect an instant output. Slow research no longer acceptable.
	Libraries are open.
	> but a push to even greater opennnes right now, and openness is tied in with collaboration. And collaborative standards
	Libraries produce outputs.
	> and what they are is changing.
	Libraries evolve.
	> Libraries do not have a great history of making rapid change - like to think long and hard - but speed now important
	Libraries provide longevity

	How to stay relevant?
	> question of how to maintain our role as custodians in a changing environment
	>> working with academics, getting involved in projects as early as possible, be open on issues, visible on cost.
	> hosting vs archiving
	> pay attention to users as well as project requirements.

	Seperate content from interface.
	Content is to be saved, interface is not built forever. Interfaces ephemeral.
	Content sits on top of metadata and infrastructure.
	[I think this is really important]

	Matthew McGrattan, Part 2: Building a digital collection.
	Digitisation Service Manager & Imaging Specialist

	The Digital Turn... access, opporunities, dissemination
	The Basics of starting a digital project... What do you have? Who is going to use it? How are you going to build it?

	What
	Content, plus what is going to sit on, around and below the content

	Who
	Specialists? Technical terminology may be off-putting depending on target audience.
	But of course also non-specialists > striking the right balance.
	Machines. Machine readability important.

	How
	Driven by what and who.
	Example of questions can be seen when we think about digitising images.
	- What format - lossy or lossless?
	- OCR needs 300pi or above.
	- What is preservable? Future-proofing and format integrity.
	- Metadata: stored within the image or alongside?
	- Display: flat, panning/zooming, comparative interface, image manipulation.

	A warning about standards!
	Standards proliferation doesn't usually give you anything extra, just more standards
	And yet we do want to avoid silos, so unification of standards is worth the effort - Digital.Bodleian project.
	Used a stack of services from infrastructure to user interface: combintion of standard services and developed tools.

	QA
	Departments territorial, libraries neutral: how does this work at the Bod?
	We can't tell anyone to do anything, so we create tools/environment for collaboration.
	We work on lots of these projects, individual scholars may be working on one for the first time.



	Javier Pereda, Interactivity and Inferfaces

	Way of trying to make us think differently about our data.
	Interaction not just between people, but also machines/systems. Reciprocal interaction between entities.
	Interfaces are not just computers...
	So think about interfaces outside of the context of the computer to create good digital interfaces.
	QWERTY Keyboard an example of an analogue interfact transposed on the digital when the analogue complexities that dictated the original design have been solved.
	Command line the origin of how we talked to computers.
	This evolved into the text menu interface: but now the BIOS is a cumbersome interface.
	Graphical User Interface. GUIs give simple or elaborate visual representations of the object.
	Studies in human computer interaction take this further...
	Natural User Interface. Natural human interfaces.
	- Multi-sensory interfaces (eg iPhone gestures).
	- Human like interaction. Make the actions similar to what a human would do.
	- Biometric interaction.
	- Invisible computing.
	Tangible interfaces
	> require less learning than a GUI.
	> user just thinks for his or herself, doesn't need to know how the interface works.

	The Internet of Things
	Allows objects to react, with or without us.
	Though codependency important: if we didn't use wikipedia, wikipedia wouldn't interact with other objects/things.
	Stop thinking of computer as the screen which presents information, rather think how best to present information.

	Need to think beyond the keyboard and the mouse and the gestures on a tablet.
	And interfaces which people aren't afraid to use, unlike the keyboard/mouse.

	Users learning from mistakes... do they want to? what kind of mistakes are acceptable/invisible?
	[this seems a bit like making every interface as fuzzy as possible given the technology available at a given moment]


	Ségolène M. Tarte, Working with Digital Images

	Why images? Conserve, preserve, cuarate historic documents.
	But also to analyse, study and interpret them & document, present and disseminate.

	What are digital images?
	x, y + 0-255 (for greyscale)
	alpha is a transparency value.
	HSL: hue, saturation and lightness on 0-1 scale > if l=0 colour is black, if l=1 colour is white, if s=0 colour is grayscale.
	Mach band effect [now that is fascinating] > shows that perception of brightness within greyscale helps us detech changes.
	So, when processing/digitising an image we are making an act of interpretation. Because seeing is subjective and image capture interpretive.

	Histogram based processing of images: counting values (say between 0 and 255 on a histogram)
	(Logarithmic count scale more useful at times than linear count scale)
	When we adjust the brightness, so we can see better, the histogram changes.
	- we take the whole thing and push it along the scale.
	When we adjust the constrast we strech out the grey values, redistribute them along the scale.
	(though usually you adjust a bit of both)

	Image segmentation

	Problem: you know what you are looking for but you need to really know what you are looking for to be able to find it...
	How then?
	Identify features within an image: such as blobs, lines, edges (things that are tripped up on along the way)
	> related to brightness persception, so greyscale images usually used for detection.
	Or region identification: so for example, an area not covered in ink (sweeping up similar things)
	Sobel filters help scratch detection.
	Thresholding: between that and that make it white, between that and that make it black.

	Processing images is modifying them, interpreting them (so keep the original!)
	Understanding the (ofte black-boxed) image processing options helps justify choice and make expectations explicit
	> the act of interpretation then becomes more shareable, reproducible, and teachable.

	GIMP
	Windows > Dockable dialogs > Histogram
	Colours > Hue > move saturation to -100 to make things greyscale.
	Colours > Desaturate
	Colours > Brightness-Contrast
	... histogram here is useful for keeping track on precisely what we are doing


	RTI Viewer

	Uses multi image capturing processed using Polynomial Texture Mapping (PTM) to display the image.
	Move light around the object to simulate depth, and simulate our looking around an object to understand it.
	Coefficient means that only the equilavent of 9 greyscale images are saved (RGB (so 3) plus 6 coefficients)

	Multi-image use in GIMP
	Blend modes... layered stack. Use Windows > Dockable Dialogs > Layers and then Mode to combine/compare images.
	Chances are that blending them (and lightening from a white background) will illuminate the text.
	Do lighten only with black bottom layer, and darken only with white bottom layer, then stack two compiled images using difference.


	*This work is licensed under a Creative Commons Attribution 3.0 Unported License.*

	<a rel="license" href="http://creativecommons.org/licenses/by/3.0/"><img alt="Creative Commons License" style="border-width:0" src="http://i.creativecommons.org/l/by/3.0/88x31.png" /></a><br />This work is licensed under a <a rel="license" href="http://creativecommons.org/licenses/by/3.0/">Creative Commons Attribution 3.0 Unported License</a>.