#361 an addition to <timeline> to record stretches of time

AMBER
closed-rejected
Lou Burnard
5
2012-09-20
2012-05-05
Sebastian Rahtz
No

The <timeLine> element allows us to talk about something which takes place over a period of time, and lets us
talk about points within that period, using <when>. However, this does not let us talk about _chunks_ of the timeline.
I'd like to use such a thing to link a text to an audio or video rendering of it (a performance of a play, for example). Thus
I have a poem with 8 lines, and 60 seconds of audio of the poet reading it. I'd like to note that line 2 can be linked to the
audio, the point between 8 seconds in and 15 seconds in. I have in the past abused <when> by adding @from/@to attributes
to it, which works fine, but it would be good to have this done cleanly. One _could_ define stretches of time by defining <when>s for the start and
the end, but this is cumbersome, and does not provide a unit to link to.

My suggestion is to add a new empty element <chunk> or <stretch> or <anothername>, which can occur as a child of <tlmeline>, freely
interspersed with <when>s, and with attributes @from, @to, and @unit.

I'd also be just as happy if <when> was re-defined to allow us to talk about stretches of time, by adding @from and @to.

Discussion

  • Laurent Romary
    Laurent Romary
    2012-05-07

    I wish we could consider this point in relation to the issue of character-shift based stand-off (is there a difference between offsets on a string and offsets on an audio channel?). For the record, I append below part of the message I sent to the stand-off markup gang, which initiated quite a discussion in the last week:

    I am currently finalizing the ISO 24611 document on Morphosyntactic annotation, where we have a proprietary notation for stand-off with character offset (see example blow) and for which I would like to present a TEI based alternative syntax. My assumption is that if TEI and ISO could speak the same language, we would probably be more likely to get some implementation (we could even communicate on this).
    http://www.tei-c.org/release/doc/tei-p5-doc/en/html/SA.html#SATSSR seems the appropriate means to do what I want here, but there is no paragraph in the guidelines (with an example) eliciting its use for stand-off mark-up. To my view, the semantic is clear here (the <w> should be used as containing the corresponding string, further reference to the source is possible (e.g. to get its metadata, etc.); but I would not want to have this discussion opend now at the expense of losing track of the request).

    So, given the example below, I would be delighted if you could provide guidance on:
    - telling me if the use of @coresp is appropriate, if not provide a stable alternative
    - telling me if the syntax for the pointer is appropriate, if not provide a s table alternative
    - considering if the guidelines should be more pushy in recommending this (I would then open a ticket accordingly)

    We really suffer from not offering a ready made character based stand-off mechanism to the NLP community. Maybe this is an opportunity to do so.

    +++++++++ Source example +++++++++++++++
    I know, it's missing a fragmentIdentifier (sigh...)
    <token from="0" to="3"/>
    <token from="4" to="9"/>
    <token from="10" to="11"/>
    <token from="13" to="19"/>

    +++++++++ TEI Encoding +++++++++++++++
    <s xml:id="s1">The victim's friends</s>
    <w corresp="#string-range(s1,0,3)"/>
    <w corresp="#string-range(s1,4,6)"/>
    <w corresp="#string-range(s1,10,2)"/>
    <w corresp="#string-range(s1,13,7)"/>

     
  • I feel fairly unable to comment on Laurent's example without spending a few hours realigning my brain to the TEI pointer business. It had not occured to me it was the same problem as the audio, but I now see that it could be.

    you _could_ turn this round and model it on <timeline>, with
    <textstream>
    <token xml:id="x1" from="0" to="3"/>
    <token xml:id="x2" from="4" to="9"/>
    <token xml:id="x3" from="10" to="11"/>
    <token xml:id="x4" from="13" to="19"/>
    </textstream>
    <w corresp="#x1"/>
    etc
    which has the virtue of being easy to process, and allows creative use of @unit
    to vary what the numbers mean.

    I find <w corresp="#string-range(s1,0,3)"/> opaque. what is the "s1" there? it claims to be
    a "fragment identifier", but what does that mean? is it a URI pointer, or what?

     
  • Laurent Romary
    Laurent Romary
    2012-05-07

    Clever proposal. That could be something to integrate.

    As to your question, this is the essence of the lengthy discussion on the TEI-SOM list....

     
  • Lou Burnard
    Lou Burnard
    2012-05-15

    As you note above, the existing mechanisms provide for the alignment of stretches by requiring you to supply <when>s for the start and for the end of the stretch. This seems much more flexible (you can for example say that you don't know exactly when the start or end was relative to the end or start), though I take the point that it may be more difficult to process without some intermediate element

    I wonder whether it would be appropriate to use the existing <span> for that. This would give us something like

    <l xml:id="line2">blah blah blah</l>
    ...
    <span from="#t1" to="#t2 sync="#line2"/>
    ...
    <when xml:id="#t1" when="000475960"/>
    <when xml:id="#t2" when="000475970"/>

    You could even cheat by using identifiers for the <when>s which included the time stamp!

    <span from="#t000475960" to="#t000475970" sync="#line2"/>

    If the semantics of <span> such as they are preclude this, then I guess it has to be a <link>

    Laurent's points seem to merit a different ticket so I don't address them here.

     
  • Yes, I could use <when> to define points in time, and then a series of <link> elements to create an intermediate object which points to the start and the end. It is pretty cumbersome though. Is there a downside to an element which just does the job inside <timeline>?

     
  • James Cummings
    James Cummings
    2012-06-29

    • assigned_to: nobody --> louburnard
     
  • Lou Burnard
    Lou Burnard
    2012-09-16

    • milestone: --> AMBER
     
  • James Cummings
    James Cummings
    2012-09-20

    • status: open --> closed-rejected
     
  • James Cummings
    James Cummings
    2012-09-20

    Council rejected this ticket at 2012-09 Oxford face 2 face, finding that the Guidelines do cope for this already. However, notes that timeline should be revised as part of a larger re-assessment of temporal markup.