OASIS Open

Find connection link within this action item

https://www.oasis-open.org/apps/org/workgroup/lexidma/members/action_item.php?action_item_id=3939

- The meeting started at 15:00 Central European Summer Time. Chair: David Filip. Minute taker: Michal Měchura.

- Attendance was taken and voting rights updated

dF, Miloš, Ilan, Simon, John, Michal, Mark, Carole

Quorum reached with 6/8 voters present

Mark gained voting rights after June meeting, Michal will gain voting rights after today's meeting

- There are no minutes from the last meeting yet. So, we will be approving minutes from two meetings at the next one.

- The chair acknowledged the two discussions that have been going on on the mailing list since the last meeting, one on multi-level senses and one on Ontolex.

- The chair asked whether there is consensus on the fused design principles. There is consensus on everything except the question of recursion (= whether we should or should not allow recursion in the data model).

https://github.com/oasis-tcs/lexidma/wiki/Fused-revised-design-principles was edited during the meeting to reflect lack of consensus on recursion.

A discussion ensued from which the following insights emerged:

-- Recursion is when an object of one type is allowed to contain other objects of the same type. Typically: when a sense is allowed to contain further senses (= subsenses). Embedding by recursion can go on into infinity, although it would appear that in practice dictionaries rarely go beyond two levels of embedding.

-- Recursion is a distracting complication for various digital agents (= software tools that process dictionary entries). Some members argue that this concern is non-trivial and should inform our design principles, given that the mission of Elexis is to make lexicographic data more easily processable by machines.

-- Recursion is sometimes allowed in dictionary schemas (including Ontolex and TEI) to model phenomena such as: (1) subsensing, as a way to group senses that somehow belong together, and (2) entry grouping, for example to include the entry for 'walk' (noun) as a "run-on" or subentry inside the entry for 'walk' (verb).

-- There is consensus within Lexidma that our standard needs to be be able to model these phenomena. But, there is no consensus on whether this means we must allow recursion on some data types. Recursion is not the only way to model these phenomena. Alternative propsals exist, several generated by ourselves within Lexidma recently, which avoid recursion:

--- For entries, the consensus in Lexidma is that there should be no "entries inside entries" but, instead, that the entries that belong together should be connected via entry-to-entry relations (yet to be elaborated by Lexidma).

--- For senses, the usual solution (in dictionary schemas where the authors wish to avoid recursion) is to establish two (or possibly more) types, 'Sense' and 'Subsense' (and possibly 'Subsubsense' and so on), and to declare that 'Sense' can contain 'Subsense', and so on. This is the same as allowing recursion but only to a specified number of levels, and the cost is a proliferation of data types (≠ "keeping it simple"). It appears that Lexidma members do not find this solution very smart.

--- An alternative proposal is to do for senses what we are already planning to do for entries: keep the list of senses flat (ie. no recursion), while the senses that belong together would be connected either via sense-to-sense relations (yet to be elaborated by Lexidma, one of which could be simply "is subsense of"), or by means of labels/tags (also yet to be elaborated).

- Action item: Michal Měchura will look at the schemas of existing dictinaries (those available through Elexis, and others) and try to come with some emprical observations on recursion (and recursion-like things): how prevalent is it, how many levels, what phenomena is it used to model, could those phenomena be modelled using something else, what would be lost or gained in such remodelling? Michal to submit this analysis in time for reactions before next meeting..

- It was agreed to move the date of the next meeting from 17 August 2020 to 24 August 2020.

- The meeting ended shortly after 16:00 Central European Summer Time.

1. Administrative

• roll call

• appoint scribe

[need a pool of scribe volunteers to rotate the duty]

• approve minutes from previous meeting
  minutes from June 15 pending, action Simon

2. Material

• DMLex Working Draft format Docbook template set up pending https://github.com/oasis-tcs/lexidma

discuss normtaive keywords style

• DMLex WD Technical discussion
  • Based on technical contributions received by 20th July 2020
    •  
• Continue discussion of design principles
  • Do we have consensue on the fused design principles?
    https://github.com/oasis-tcs/lexidma/wiki/Fused-revised-design-principles
• Continue discussion of  data model diagrams
  • New contribs
    • Miloš, Ilan, and John discussiong "two level senses"
      https://markmail.org/thread/ysxlq3ih5r4q6b67
    • John and Michal discussing revised Ontolex diagaram
      https://markmail.org/thread/wmmarqw5ugekxvdo
  • Miloš: https://www.oasis-open.org/committees/download.php/66840/LEXIDMA_ER_version1.pd
  • John: https://github.com/oasis-tcs/lexidma/wiki/Data-Model-based-on-OntoLex-TEI-Lex0

Result of clarification: Link exists at Entry level, has source and target selector, also an attribute for type of relationship

a possible @POS data model, instead of free form as proposed by Miloš

lex:NMTOKEN

[authority]:NMTOKEN

possible authority prefix examples:

[universal part of speech tagging]

upos:

clarin:

lemon:

[how to relate (inflectional) forms to POS?]

Morphology object with possible value pairs

@POS="[authority]:NMTOKEN" 

@formsLink="[formsURI]"

example: @relationshipType="imperative"

John's and Miloš's diagrams serve as far points that need to be brought together before spec writing can start

the situation is really a triangle with Lemon<>Lex-0 zero being more fine grained, Miloba and Lemon being more prescriptive in allowed relationships -: better for industry interoperability

• LEXIDMA TC liaisons and appointement of reps
  • TEI Lex-0 appointement of rep pending

Call for a liaison rep 

3. Closing Admin

• Recap of agreed actions
• AOB
• Adjourn [no later than 1400 UTC]

• OASIS Open (General Membership)
• General Public