Relations that are seldom or never signaled

[Maite Taboada]

Dear all,

My two cents on the signaling discussion:

I have analyzed a corpus of 60 task-oriented conversations (collected by
the ISL at Carnegie Mellon University), in English and Spanish (30
conversations in each language). These are the percentages of relations
that are signaled by a discourse marker:

English: 30.86%
Spanish: 45.88%

Relations that are never or rarely marked:
Elaboration
Enablement
Evaluation
Evidence
Motivation
Restatement
Solutionhood

This "marking" includes only traditional conjunctions, not even things
like "well", "I mean". I think we need to take into account what are we
considering as connectives, because that's also going to change the
results we get. As was mentioned earlier in this thread, there are many
other devices to signal a relation. For instance, most of my Solutionhoods
were questions in a broad sense (not a strict question-reply pair, but
something where the answer to the question might be the rest of the
conversation, as in "When would you like to meet?", which prompts a whole
dialogue on arrangements for an appointment). Then, interrogative mood
"signals" (one type of) Solutionhood, but we're not including that in our
count.

Spanish and English were quite comparable in signaling, except for Causes
and Results, which were signaled much more in Spanish, the Volitional ones
actually were signaled whenever they were used in the Spanish data.

The most interesting thing, however, was that my corpus was spoken,
dialogic, language. Leaving aside "holistic elements", that is, greetings
and the sort, I could produce a full RST tree for every single
conversation. Now, the conversations were task-oriented and quite
"focused" because of that, I'm not sure how spontaneous conversation would
be analyzed. For instance, I don't know how to deal with interruptions
that do not contribute to the ongoing discussion (interruptions that are
not clarifications). I'm also interested in whether there are different
types or frecuencies of signaling in spoken versus written language. But I
haven't gotten there yet. So, I continue working, since I still have
unanswered questions...

Cheers,

- Maite


          		Maite Taboada
Department of Computing Science AND Modern Languages and Cultural Studies
615 General Services Building     Tel: (780) 452-0314
University of Alberta             Fax: (780) 492-1071
Edmonton, Alberta                 e-mail: mtaboada at ualberta.ca
T6G 2H1 Canada                  http://www.ualberta.ca/~mtaboada