next up previous contents
Next: The Data Up: {eventive verb + NP} Previous: {eventive verb + NP}

The Analysis: Methodology and Assumptions

The first phase of the corpus analysis involved extracting all sentences containing begin and finish in any of their inflected forms. I did this in the case of the LOB simply by utilising regular expression matching commands standard under UNIX. The BNC demanded a more advanced mechanism due to its size and structure, and hence I used the CORSET corpus search toolkit (Corley 1996) . This tool enabled extraction of only those sentences which do not have an explicit VP complement. Further narrowing of the sentences to be examined was accomplished via regular expression matching, picking out only those sentences in which the aspectual verbs were followed by a noun phrase. At the end of this phase, I had a large collection of sentences which were potentially metonymic, as each sentence contained an aspectual verb followed by a non-VP element.

The second phase involved identifying which of the collected sentences were actually metonymic. As neither of the two corpora (or indeed any corpus to which I have access) contains any kind of semantic tagging, much of this work had to be done by hand. The following cases were eliminated from the sets of sentences:

The third and final phase was to read through each of the metonymic sentences in context in order to determine the interpretation intended by the speaker/author. In addition, the whole procedure was repeated to find metonymic instances of the phrase begin on followed by a noun phrase for purposes of comparison with begin. The results of these analyses will be introduced in the next section. They provide the basis for the alternate conception of logical metonymy to be put forth in section 5.5.2.

 

Begin Finish Begin on
Agentive
write WRITTEN_OBJECT 34 = 20.7% 44 = 13.8% 0 = 0%
say sentence 4 = 2.4% 7 = 2.2% 0 = 0%
have family 1 = 0.6% 0 = 0% 0 = 0%

OTHER AGENTIVE 26 = 15.9% 43 = 13.5% 4 = 16%
Subtotal Agentive 65 = 39.6% 94 = 29.5% 4 = 16%
Telic
eat FOOD/MEAL 11 = 6.7% 87 = 27.3% 1 = 4%

drink LIQUID 0 = 0% 55 = 17.2% 0 = 0%

tell STORY 20 = 12.2% 7 = 2.2% 0 = 0%
attend school/classes 4 = 2.4% 22 = 6.9% 0 = 0%
sing SONG 9 = 5.5% 6 = 1.9% 0 = 0%

play MUSIC 7.5 = 4.6% 2 = 0.6% 1 = 4%
read WRITTEN_OBJECT 2.5 = 1.5% 10 = 3.1% 0 = 0%
serve (jail) sentence 8 = 4.9% 0 = 0% 0 = 0%

smoke cigarette 0 = 0% 5 = 1.6% 0 = 0%

do business 6 = 3.7% 1 = 0.3% 0 = 0%

take MEDICINE/TREATMENT 5 = 3.0% 1 = 0.3% 2 = 8%
do DEGREE 1 = 0.6% 6 = 1.9% 0 = 0%

do homework 0 = 0% 5 = 1.6% 0 = 0%
play GAME 4 = 2.4% 0 = 0% 0 = 0%

argue CASE 3 = 1.8% 0 = 0% 0 = 0%

present EVIDENCE 1 = 0.6% 2 = 0.6% 0 = 0%
OTHER TELIC 9 = 5.5% 2 = 0.6% 1 = 4%
Subtotal Telic 91 = 55.4% 211 = 66.1% 5 = 20%
Context 8 = 4.9% 14 = 4.4% 16 = 64%
Table 5.1: Distribution of Logical Metonymies for begin, finish, and begin on in the BNC

 


next up previous contents
Next: The Data Up: {eventive verb + NP} Previous: {eventive verb + NP}