Welcome, Guest. Please Login
Tinderbox
  News:
IMPORTANT MESSAGE! This forum has now been replaced by a new forum at http://forum.eastgate.com and no further posting or member registration is allowed. The forum is still accessible via read-only access for reference purposes. If you wish to discuss content here, please use the new forum. N.B. - posting in the new forum requires a fresh registration in the new forum (sorry - member data can't be ported).
  HomeHelpSearchLogin  
 
Pages: 1 2 3 4 
Send Topic Print
Tinderbox for Textual Analysis (Read 110411 times)
Mark Bernstein
YaBB Administrator
*
Offline

designer of
Tinderbox

Posts: 2871
Eastgate Systems, Inc.
Re: Tinderbox for Textual Analysis
Reply #15 - Jun 26th, 2009, 11:22am
 
For lead references on qualitative analysis in sociology, see
http://caqdas.soc.surrey.ac.uk/.

Lots of relevant action at Digital Humanities 09 last week, not to mention the Joint Conference on Digital Libraries.
Back to top
 
« Last Edit: Jun 26th, 2009, 11:23am by Mark Bernstein »  
WWW   IP Logged
Loryn
Full Member
*
Offline



Posts: 97

Re: Tinderbox for Textual Analysis
Reply #16 - Jun 27th, 2009, 10:28pm
 
Ooh, thank you Mark, that's pretty helpful.

I've just posted the final installment of the narration of my first Textual Analysis example: http://loryn.me/journal/2009/6/28/discovering-poetry.html

The final step is to examine the interaction between available affordances in my tool chain, and to see how well my original post in this thread holds up against an analysis of the detailed recounting of this particular analysis.

I think it'd be instructive for me to proceed to narrate in a similar fashion:
- a knowledge cafe-generated corpus analysis
- multiple linguistic analyses of a short story
- a qualitative sociological study
- a semiotic image analysis

Each narration concluding with an examination of the affordances available within the tool-chain, and a projection that examines the likely impact, and design of, new affordances.

(Someone stop me if they can think of a short-cut to the end-game. This is hard work!)

If there's any Sydney-based lawyer reading this, I'd be keen to examine your practices surrounding textual analysis.

As for the rest of my day, I guess I'll be reading A Companion to Digital Humanities ( http://tr.im/pNIe ).
Back to top
 
 
  IP Logged
Mark Anderson
YaBB Administrator
*
Offline

User - not staff!

Posts: 5689
Southsea, UK
Re: Tinderbox for Textual Analysis
Reply #17 - Jun 28th, 2009, 6:57am
 
Looking at the last part of your article 'Reflowing text', I'd note that you can actually link words - text links can target text within a note as well as just the note itself. However the utility of this is limited as I've never found a means to show text anchors for inbound links. so whilst no that usable today, it is an existing feature that, with some polishing might help towards the sort of analysis you're describing.
Back to top
 
 

--
Mark Anderson
TB user and Wiki Gardener
aTbRef v6
(TB consulting - email me)
WWW shoantel   IP Logged
Loryn
Full Member
*
Offline



Posts: 97

Re: Tinderbox for Textual Analysis
Reply #18 - Jul 1st, 2009, 8:35am
 
Hi MarkA

I did comment on using text links in the article Reflowing Text:

Quote:
Right now in Tinderbox, I can only choose to create named semantic relations between notes; or if I used text links, to have an unnamed relation, and then, to be able to make only one link, not several, as so often happens in analysis.


Although I hadn't considered the inability to create text anchors. Thanks for pointing that out.
Back to top
 
« Last Edit: Jul 1st, 2009, 8:36am by Loryn »  
  IP Logged
Loryn
Full Member
*
Offline



Posts: 97

Re: Tinderbox for Textual Analysis
Reply #19 - Jul 1st, 2009, 8:42am
 
I've published my conclusion to the Textual Analysis Example 1 series. Affordance Critique ( http://loryn.me/journal/2009/7/1/affordance-critique.html ) extends the original post in this forum thread.

I've stored the original call for extending Tinderbox, at http://loryn.me/tatbxmanifesto/

I've stored the Textual Analysis Example 1 series at: http://loryn.me/taeg1/

And I've stored the complete set of writings on this topic, including all textual examples and discussions about extending tinderbox functionality at: http://loryn.me/research/
Back to top
 
 
  IP Logged
Loryn
Full Member
*
Offline



Posts: 97

Re: Tinderbox for Textual Analysis
Reply #20 - Jul 1st, 2009, 8:45am
 
Did you all see John Stephan's note?

Quote:
Loryn: I couldn't agree more. This would be a fantastic addition to Tinderbox and greatly enhance my ability to use it in my law practice.


It seems like there's some demand for functionality along these lines.

Feel free to leave your own comments about this possibility, either here, or on http://loryn.me
Back to top
 
 
  IP Logged
Mark Anderson
YaBB Administrator
*
Offline

User - not staff!

Posts: 5689
Southsea, UK
Re: Tinderbox for Textual Analysis
Reply #21 - Jul 1st, 2009, 10:17am
 
Thanks for an interesting series of articles.  If I may, I'd just note a couple of erros inrealting to current Tinderbox functionality. So:

Quote:
Right now in Tinderbox, I can only choose to create named semantic relations between notes; or if I used text links, to have an unnamed relation, and then, to be able to make only one link, not several, as so often happens in analysis.


Not strictly true!  A note can have have more than one link from the same source, be it at note level or a text selection (your 'syntagem lense'?).  The links can go to different targets.  It is also incorrect to say relationships are 'unnamed'.  They'll only be that way if you don't give the link at name.  When you create a link you set the type; leave blank, pick from the list or add a new link title (which is subsequently a list item). Within the current app session, for subsequent links type defaults to the types used for the last link created in the session, reverting to nothing (no title) at the start of a new session. The action codes like linkTo(item[,linkType]) that create/remove links also have an optional argument for a link type (label). The Browse Links dialog also shows - for each outbound link present: the target note, the source link text (if any) and the link type name. I'd agree there are issues visualising all this in some other contexts in the app but that's a different aspect that I'll touch on further below.

In you last part of the article (affordance-critique.html) it says: Quote:
it is more difficult to read the text as a flowing whole, especially where the verse division occurs within a clause complex

Have you fully investigated use of Nakakoji view?  With appropriate scope/templates it is possible so show a section of text made up from many notes.  In this context the view is better thuoght of as a "Text Viewer view" rather than a "Text Export view". The view can be used for viewing text - you don't have to export anything.  By resizing the window and setting the appropriate preferences/attributes to use a larger type size you can easily read sections of text that might be from (non) contiguous notes (a 'syntagem lense' view?) and/or even include other thinks like attribute values.

Admittedly, my corrections don't make massive difference to your inability to do your process in Tinderbox but it does help establish what parts of it can't be done. so this isn't too long i'll pick up some other points in another reply.
Back to top
 
 

--
Mark Anderson
TB user and Wiki Gardener
aTbRef v6
(TB consulting - email me)
WWW shoantel   IP Logged
Mark Anderson
YaBB Administrator
*
Offline

User - not staff!

Posts: 5689
Southsea, UK
Re: Tinderbox for Textual Analysis
Reply #22 - Jul 1st, 2009, 11:14am
 
Despite what you can do with links, I'd agree it is difficult in some contexts do some of the things you task (and others need):
  • The exact in/out points for links source anchors are hidden in the TBX source so wokring with overlapping links - whether making or maintaining them isn't easy.
  • To use a text target anchor for a link you must create a selection in the target note (i.e. before dragging out the link) and also action codes like linkTo() don't suport link text target anchors so automiation becomes less easy.
  • When following links with text target anchors the only difference from links to notes is that if the text is long and the target is not immediately on view, the text pane is scrolled to display the target
I don't like to think of the above as a failure of TB though as none of the above were - as far as I know - original design assumptions. So, despite the app's adaptability, the fact it doesn't quite stretch far enough is a frustration not an error.
Which leads to 'feature creep'. If you look at the changes to Roadmap view over the life of TB it's acquired numerous extra feature enhancements. Only a few days it was pointed out that one can't - in Roadmap - see different link type labels. Given the design that's not surprising.  The question becomes not just can the desired new enhancement be added, but is it sensible to do so. In reading the fine detail of parts on the needed enhancements for your analytical process, I do find myself thinking whether grafting this onto Tinderbox is the right way to start or whether you need a different tool, very like Tinderbox in some aspects, but where the app is more focussed on the quite complicated and particular needs of the text analysis method being proposed. On another facet how much performance hit will there be for the many users who won't use the posited additions? It's all a matter balance.

Parking the latter issue, and returning to what we can/might do, I think a concordance view might not be that hard to do. Consider the 'Find' view. We give TB a query (in Find simple but it could be more open-ended like AgentQuery) and we get a list of matches.  The list uses the $Name of matched notes, but that could could be the matched term preceded/followed by some context. The latter could perhaps be set via a $ContextStart & $ContextEnd stored expression rather like the $DisplayExpression method works. The listing would also need to no de-dupe the matched notes lists as one note might have more than one match each of which is pertinent in a concordance context. As at present, double-clicking a list item (or using its context menu) would give access to the matching note which perhaps the addition of highlighting all matches (in a way Find sort of does). I hasten to add this concordance view idea is me thinking out loud - it's just an idea at this point.

Time to break to a new reply...
Back to top
 
 

--
Mark Anderson
TB user and Wiki Gardener
aTbRef v6
(TB consulting - email me)
WWW shoantel   IP Logged
Mark Anderson
YaBB Administrator
*
Offline

User - not staff!

Posts: 5689
Southsea, UK
Re: Tinderbox for Textual Analysis
Reply #23 - Jul 1st, 2009, 11:39am
 
You also mention the problem of mixing source text with annotations such that agents get mixed results. That's unsurprising but the way around is simple - more segmentation. There are 2 methods I can think of.

The footnote tool. I believe this feature was made for exactly this problem of avoiding source and comment in the same (atomic) note. See the PDF manual page 103 for more on this. A starting link for aTbRef on the subject is here. The footnote tool automatically makes mutual links with discrete labels: 'note' for footnote to note and 'note+' for note to footnote. The footnotes can either be a sibling of their 'source' or a child.

Use $Name for the source text and $Text for comment. Changes in v4.6.x relating to showing long titles and changes in v4.7.0 relating to Exploding notes and conditional expressions in Display Expressions. So, referring to your project, we'd store '1 John 1:4' in $SourceReference (attribute name for example only)**. $Name is "And we are writing these things so our joy may be complete.". Our Display Expression will be something like
  $DisplayExpression = $SourceReference + ": " $Name
or
  $DisplayExpression = $Name + " (Ref:" + $SourceReference + ")"

As the source location is in an attribute we can sort on it (where sort is needed) even though on screen the data's maybe only seen as part of a Display Expression.

"But wait," you say, "when I'm editing the text window I don't see the key thing - the source text. Just ensure $ShowTitle is set to true. Conveniently, this shows just the $Name value and not the $DisplayExpression value.

** this setting up of attribute values can be done as part of data import/explode, with a little planning. I've made a post today in the Q&A forum pointing to some updated info on use of explode.

OK, none of this is everything you need but hopefully, I've made a few things a little easier. Now an empty tea cup and a burgeoning in-tray beckon...
Back to top
 
 

--
Mark Anderson
TB user and Wiki Gardener
aTbRef v6
(TB consulting - email me)
WWW shoantel   IP Logged
Loryn
Full Member
*
Offline



Posts: 97

Re: Tinderbox for Textual Analysis
Reply #24 - Jul 1st, 2009, 4:13pm
 
Aha. Indeed you have made some things easier.

I do hope no one reads my set of articles as any sort of indictment on Tinderbox. I am consciously examining a new set of usage scenarios for Tinderbox, a set for which Tinderbox was not initially designed.

My initial insight (http://loryn.me/tatbxmanifesto ) was that there are a minimal set of facilities that would fit reasonably closely with the existing machinery and philosophy of Tinderbox that can go a long way in providing the functionality people need to be more productive in performing textual analysis.

Now in my article series, I am definitely exploring some more esoteric features, because I'm taking my starting point as the practice of textual analysis, and my question is, "What would Tinderbox (or some other tool) need in order to completely provide for my need in this practice?" Of course, subsequent to that----and I've already published my intention (on the first page of this forum)---I'll then turn to the question as to, "Do these minimal facilities that I've discussed bring the bang-for-buck that I supposed they would? Did I describe them accurately in the first place? Can I flesh out a design that would significantly enhance Tinderbox, fit well with its machinery and philosophy, and still provide advantages to most users? Are they general-purpose enough to be worth adding to a general-purpose tool like Tinderbox?"

So, you're absolutely right about the fact that Tinderbox most probably won't ever accommodate *everything* I set out to highlight in this initial series. But I am hopeful that this analysis, and whatever papers flow from it, do influence Mark's planning for Tinderbox 5 in some way.

And, MarkA: Thank you for correcting me about my misunderstanding of Tinderbox link functionality, and for expanding my understanding of the existing affordances within Tinderbox for what I'm seeking to achieve.

I think that finding, "Well, Tinderbox can already do this non-obvious task," is an extremely positive outcome! I think your comment re: reference source, etc. is right on the money in this regard.

wrt: Nakakoji view. I'm aware of it. But it is *read-only*. I need a view that allows further analytical work to be done.
Back to top
 
« Last Edit: Jul 1st, 2009, 4:15pm by Loryn »  
  IP Logged
Mark Anderson
YaBB Administrator
*
Offline

User - not staff!

Posts: 5689
Southsea, UK
Re: Tinderbox for Textual Analysis
Reply #25 - Jul 1st, 2009, 4:46pm
 
Loryn, thanks.  I'd hope on one would mis-read your intent. I didn't read is as knocking. The comment about whether it's the wrong place to start (from existing TB) is less a matter of knocking your concept and more just asking whether the best way to achieve the aim might be to branch the app. I believe parts of TB and Storyspace are shred.

Don't dismiss your needs as 'esoteric'. One of the things I've taken away from the TB weekends I've been fortunate enough to attend is that many TB users age here because other apps fare less well with their needs. The analytic needs your describe aren't unreal but I suspect (I'm not a programmer) some of the visualisation aspects are more complex than at first sight.

On Nakakoji, I'm interested to hear - assuming it's showing the text 'span' you 'd want - what is/are the function(s) you'd want to do based on that text.

From my non-coders understanding of XML I believe some of the things proposed would test the way some things are stored at present. Not a reason not to do things, but part of the reason for my observation about is this the best baseline to start from. IOW, starting from a differewnt standpoint/storing some things a little differently might make the overall task achievable if not easier.

Regardless, I've found this thread - and your linked articles - very interesting.
Back to top
 
 

--
Mark Anderson
TB user and Wiki Gardener
aTbRef v6
(TB consulting - email me)
WWW shoantel   IP Logged
Jean Goodwin
Full Member
*
Offline



Posts: 136
North Carolina
Re: Tinderbox for Textual Analysis
Reply #26 - Jul 2nd, 2009, 8:23am
 
Morning, all:  Thanks for this interesting discussion!  Here's my perspective, FWIW.

Since there's special purpose (and now often free) software available for many "textual analysis" tasks, I confess I hadn't even thought about turning to Tinderbox for them.  For example, antconc (http://www.antlab.sci.waseda.ac.jp/software.html) is a fantastic (free) concordancer, and Charles Turner recently pointed us to what looks like a (free) tagging system (TAMS, http://www.eastgate.com/Tinderbox/forum//YaBB.pl?num=1241539776/15#17).  

I also worry about feature bloat in Tinderbox, since I'm not using all of the features I know about, much less the many that I've not yet discovered....

Still, since we've talked about tagging (a/k/a "coding") texts twice now within a couple of months, it would be cool if Tinderbox supported this!  (I mean, supported it for lazy users like me, without the programming-knowledge that Charles has brought to the task.)

Tagging could be like the footnote tool, with these additions:
  • When creating a tag, a dialogue box would allow me to set the link (i.e., tag) type.
  • The tags could be made visible in the "source" note--somehow!  In other words, different type tags would have to look different.  
  • (optional) I could specify which if any tags to display in the "source" note.

Beyond this, I'm pretty sure that a lot of the things I'd want to do with the data could be automated in Tinderbox as it is.

For example, an agent could assign the "comment" notes  prototypes based on their link types, and the prototypes could be set up to harvest attributes (e.g., a citation from the "source" note, or a line number from the "comment" note), and to display the info I want in the way I want (as per Mark A.'s comments).  

A primitive kind of the "overlay" Loryn talked about could be done by creating an agent to collect the "source" note plus all "comment" notes of a certain type, and then freezing the agent to allow the notes to be played with a bit.

Or the data could be examined in other ways.  Following up Victor's suggestions:  "comments" of the same type could be compared across "sources";  the "comment" notes for each "source" could be sorted by line number, and the order of the tag-types compared across "sources";  etc.

OK, I would still probably turn to special-purpose software for larger projects.  Tinderbox doesn't need to do everything.  But a tagging/coding functionality would be really useful for smaller ones, and for general brainstorming.
Back to top
 
 
  IP Logged
Mark Anderson
YaBB Administrator
*
Offline

User - not staff!

Posts: 5689
Southsea, UK
Re: Tinderbox for Textual Analysis
Reply #27 - Jul 2nd, 2009, 6:07pm
 
I may be misunderstanding, but I assume the term 'tags' in this context applies to a form of multi-value field/attribute?  Like keywords with photos?  If so, that's already supported by use of 'set' type attributes (see more).

You can add or remove a set value through action code: stamps, moving onto adornments, into containers, via agents, etc.
Back to top
 
« Last Edit: Jul 3rd, 2009, 5:19am by Mark Anderson »  

--
Mark Anderson
TB user and Wiki Gardener
aTbRef v6
(TB consulting - email me)
WWW shoantel   IP Logged
Charles Turner
Full Member
*
Offline



Posts: 180
New York, USA
Re: Tinderbox for Textual Analysis
Reply #28 - Jul 3rd, 2009, 8:14am
 
Hey Mark A.-

I think we're getting terminology clash here, that very overloaded word "tag" especially.

I think Jean is, or at least I was, referring to the attachment of attributes to chunks of text within a note. Then being able to do Tbox-like things with the attributed chunks.

I've had some interesting experience with my test implementation for Tbox. I found that there's a need to make a typographical distinction between the tag text and the note text. A tagged text can get very unreadable, very fast. The TAMS system uses text color to establish multiple channels of text in one chunk.

Also, with the work I was doing, I found that there was a real need for multiply-valued attributes to account for composers, performers and so on in each text. Given the readability issue above, I've scaled back my use of text tagging to just the note author. Tags refer implicitly to that voice.

Hi Jean- Thanks so much for the pointer to AntConc! I'm not very grounded in textual analysis, but KWIC indexes are one of my favorite ways of looking at texts. I expect to have some fun with it.

Best, Charles
Back to top
 
 
WWW   IP Logged
Jean Goodwin
Full Member
*
Offline



Posts: 136
North Carolina
Re: Tinderbox for Textual Analysis
Reply #29 - Jul 3rd, 2009, 10:34am
 
Charles is right--there's no neutral/universal vocabulary for what we're talking about:  "the attachment of attributes to chunks of text within a note"--the note being a "source," i.e. some fairly extended discourse someone wants to analyze.  Sorry for any confusion!

One of the reasons that this is hard-ish in Tinderbox is because the source text is extended:  it's important for the analyst to be able to look at it whole.  Otherwise the source could probably just be decomposed into chunks, and the chunks managed in various ways that Tinderbox already supports very well.

Here's an example:

Quote:
Four score and seven years ago our fathers brought forth, upon this continent, a new nation, conceived in Liberty, and dedicated to the proposition that all men are created equal.


(Mark A., just in case you're not following the history of your revolting colonies, that's Lincoln's Gettysburg Address._)

To figure out how this discourse "works," in my field I might want to identify all time references, all space/place references, all references to persons (including pronouns), and active versus passive sentence constructions.  So I'd want to "tag" or "code" the source something like this:

Quote:
<sentence-active><time-past>Four score and seven years ago</time-past> <person-ancestors><person-us>our</person-us> fathers</person-ancestors> <time-past>brought forth,</time-past> upon <space-continent>this continent,</space-continent> a new nation, conceived in Liberty, and dedicated to the proposition that all men are created equal.</sentence-active>


(Hey--it looks like I should be tagging metaphors, too:  "fathers/brought forth/conceived" are all in a family/birth cluster!)

After this tagging/coding process is done, I'd want (e.g.) to pull out all the time references, and see if I can see any patterns (in fact, this source moves past-->present-->future), and the same with space and person;  I'd want to see if there's any relationship between the space, time and person references;  and maybe compare any patterns in this source with other "similar" sources.  Etc.

I was thinking that I could do this in Tinderbox, IF:

1.  I could easily create something like "footnotes" to the source text--but footnotes with different link (tag/code) types (here, "space", "time", "person").  Then I'd also need to be able to do things like assign prototypes based on incoming link type, and use the place (line number?) of the target text in the source as an attribute in the child.

2.  I could get some visual indication in the source text of the material tagged/coded.  Like all "space" references have a red underline, all "time" references have green, etc.

I've been thinking about requirement #1 as using link types to represent the tags/codes being applied to a chunk of the source text.  As far as I know, Tinderbox currently doesn't give direct access to link types as a searchable/manipulatable attribute--maybe someone more knowledgeable can correct me?  Anyhow, it might be useful for many purposes other than textual analysis to be able to search for all notes with an incoming link of type x, and assign them prototype y, or vice versa.

On the other hand, it now occurs to me that something like requirement #1 could be achieved by using the footnote tool and prototype assignment.  I could "footnote" a stretch of the source ("Four score and seven years"), and then assign the child note the prototype •TimeReference, and then assign a subtag/subcode attribute "past".  I have a small analytic project coming up, so maybe I'll try it!

That leaves requirement #2--probably the bigger problem.  Since tags/codes can be embedded and overlapping and have subtypes, this is, as Charles notes, going to get complicated fast!  

Charles, thanks for the update, and best wishes on your big project!
Back to top
 
 
  IP Logged
Pages: 1 2 3 4 
Send Topic Print