"on your way to or from" (a place)

[nschneid]

@ida-szubert identified this less-than-ideal consensus annotation:

[1] consensus wb.eng_0003.56 (snt. 56 in workset wb-eng-0003, last updated on Mon Oct 15, 2012) when was the last time you were stuck in traffic that you were n't on your way to or from some sort of business ?

(s / stick-01 
  :ARG1 (y / you) 
  :ARG2 (t / traffic) 
  :time (a / amr-unknown 
          :mod (l2 / last) 
          :time-of (o / or 
                     :op1 (b2 / be-destined-for-91 
                            :ARG1 y 
                            :ARG2 (b / business 
                                    :mod (s2 / sort 
                                           :mod (s3 / some)))) 
                     :op2 (b3 / be-from-91 
                            :ARG1 y 
                            :ARG2 b) 
                     :polarity -))) 

There are a couple of interesting points here:

1. The official reification of :source is be-from-91. It seems to me, though, that there are two rather different interpretations: being from a place of origin (stative) and being in transit from a place (dynamic). Should we recognize a distinct frame for the latter case, e.g., be-coming-from-91?
2. For this sentence, we presumably will introduce a frame for the way construction, removing the need to reify :destination and :source. As currently framed, there is an ARG1 for "path, goal". Should there be an additional core argument for the source?

[kevincrawfordknight]

1) proposal

one relation (:initial-location = :be-coming-from-91)

• begin-point of some path

  separate relation for non-location uses (:source = :be-from-91)

• X born/raised in Y
• X ethnically Y
• X citizen of Y (or both)
• X competing for Y

• person from the party, information from the computer

  would require retrofitting would want to retrofit :mod? (i.e., add to "non-mod" N-N compound list) may need to! annotators will naturally shift to :source for "Romanian gymnast" anyway

  1000+ ":source" in data

2) proposal

need "way" frame? or switch to "go", which can capture both source and destination "route to the market" (no "go"?) "deported to the UK" (no "go"?) previously decided not to introduce new concepts if not required

[nschneid]

Several issues came up in our discussion:

1. If "way" (for a process of literal motion) is present, does it deserve a frame?

2. If the preposition "from" is used and no core role is available from a lexical frame, what do we do?

   1. Current policy: :source. But it's awfully vague. Among goals (in the VerbNet sense), we distinguish :destination from :purpose. Perhaps we should introduce :initial-location/be-coming-from-91 for the physical motion cases and reserve :source for more abstract starting points, such as:

      • information source: "I learned it from him/the book"
      • source of acquaintance: "I know him from math class"
      • place of historical origin: "I am from Spain [but have been living in France for 10 years]"
      • fictive motion: "a route from the market"

      (See also: FROM in PrepWiki.) A release search reveals ~1400 AMRs with :source, so this wouldn't be impossible to retrofit.

   2. An alternative would be to introduce a frame like go-01 or move-01. But according to @cbonial, we decided in reference to constructions that it's preferable to use non-core roles instead of introducing implicit frames. And there are cases where introducing go-01 feels odd for destinations ("route to the market", "deported to the UK").
3. Should we try to canonicalize modification relationships with an implicit semantics of source? "Spanish person" (i.e. person legally, geographically, ethnically, etc. associated on a long-term basis with Spain): person :mod Spain or person :source Spain?
   1. Might be difficult to define boundaries of :source in cases without an explicit "from". (Is a Toyota car a car from Toyota?)
   2. Would be nice in principle to reduce use of :mod. OTOH, retrofitting would be painful.
   3. Perhaps :mod should be tackled with a more general effort that would draw on the literature on nominal compounds from Tratz and others.