Beruflich Dokumente
Kultur Dokumente
168
13.
The first problem that should be solvedin the process of using a corpus-based
approach to theinvestigation of metaphorsis that of extracting andidentifying the
data from the corpus. Extracting relevant material from a corpus becomes extremely
difficult when applied to metaphors, as the wayin which the process of conceptual
mapping takes placeis not tied to some particular linguistic forms. Computer pro-
grams can organize language data swiftly and accurately on orthographic principles,
butidentifying and describing features such as grammatical patterns, meaning, and
pragmatic use can only be done by a human analyst [1, 92]. So a researcher whoisin-
vestigating particular linguistic units or patterns has to look through a considerable
amount of linguistic material, searching for some definite manifestations of the pat-
ternsin question. The process for extracting andidentifying relevant datain a corpus
should be carried out by the following procedures:
1.manual searching;
2.searching for thevocabulary of a source domain;
3.searching for thevocabulary of target domain;
4.searching for sentences containingitems from both the source and target domains;
5.extraction from a corpus annotated for semantic fields;
6.extraction form a corpus annotated for conceptual mappings;
7.searching for metaphors based on markers of metaphors.
Manual search consists of starting from a small corpus, or from a small part of
an already existing corpus, and searchingit manually, marking out all the metaphors
one comes across. Then, one proceeds with a larger corpus, searching for the marked
metaphorsinit. Thisis a rather efficient method, asit offers the possibility of reading a
small corpus, or a part of a corpus, entirely and thoroughly, toidentify all the existing
metaphorsinit, and by searching for themin a large corpus to receive more general-
ized linguistic results. On the other hand, this method of retrieving relevant material
limits the potential size of a corpus to a great extent.
As for extraction from a corpus annotated for semantic fields, this procedureim-
plies searching for particular linguisticitemsin the source and target domains that
have previously been taggedin the corpus. Using this method, a researcher can specify
a particular source domain, and analyze all the lexicalitems related toit, instead of
manually searching for lists of lexical expressionsa somewhat tedious and often frus-
trating process that usually yieldsincomplete lists. The analysis of target domains, and
the search for sentences containing both the potential source and target domains can
be carried out as well. As already mentioned, the main drawback of this strategyis
the rare availability of annotated corpora. Other disadvantagesinclude the fact that a
researcher must then rely almost exclusively, in hisinvestigations, on previously exist-
ing annotations. Also, semantically annotated corpora might notinclude the relevant
semantic fields for a particular piece of research.
169
Section 13. Philology and linguistics
170
13.
Searching for thevocabulary of a source domain. The first step for this type ofin-
vestigationinvolves finding the existing linguisticitems (metaphors), or whole sets of
suchitems, which represent a particular conceptual metaphor. A listis made, then a
searchis carried out with a concordance program to seeif theitems are presentin the
corpus, as the computer cannot work from a list of conceptual metaphors toidentify
their linguistic realizations [1, 93]. The selection of these linguisticitems can be based
on hypothetical decisions, on already existing lists of suchitems, or on a preceding
analysis of the keywords of the texts connected with the target-domain topics. Once
metaphors have been retrieved from a corpus, they can be further classifiedinto sub-
groups and sub-types.
Searching for thevocabulary of target domain. Different scholars have suggested
different ways of working with target domains for retrieving relevant data. In his meth-
od, whichis based on searching keywordsin the target domain, Partington [3] suggests
creating lists of terms characteristic of particular genres of discourse, analyzing them,
then running a concordance foritems that appearin more than one key list, or which
seem to belong to the same semantic set. According to Partington, such an analysis
would help reveal some systematic metaphors for experimentation, and to distinguish
the particular cases of their use. This method hasits strong and weak points. Its main
weakness liesin the fact that, for such an analysis, one needs a huge amount of homo-
geneous monothematic texts, connected to the target domain. Another weak pointis
that, in order to become a keyword expression, a word should be widely representedin
the target domain, and, thus, this type of analysis will reveal only those source domains
that are widely represented by the keywordsin the target domain. Thus, the method
will not provide utterly reliable results.
Searching for sentences containingitems from both the source and target do-
mains. Closely related to the previous two strategies searching forvocabulary from
the source or target domains is the method thatinvolves searching for sentences
containing theitems from both the source and target domains. Using this method, a
researcher should look for sentencesincluding thevocabulary from both source and
target domains. This method requires exhaustive lists of source and target domain
expressions. But to eliminate errors caused by the literal use of linguisticitems, bothin
the source and target domains, a fair amount of manual processing and editingis re-
quired, which can be burdensome. Also, itis rarely possible to find complete lists of
target and source domain expressions, so some units are likely to be missing and the
results of theinvestigation cannot be foolproof. Furthermore, this method can only
be used to reveal conceptual mappings, andis thus restricted to the metaphorical
expressions known beforehand. The main advantage of this methodis thatit can be
used for the analysis of large numbers of texts, and that, asitis based on an annotated
corpus, it can be processed automatically.
171
Section 13. Philology and linguistics
All the strategies used to extract relevant material from a corpus whileinvestigat-
ing conceptual mappings have their advantages and drawbacks. None can give com-
plete and reliable results, since all depend on the quality of the software used and on
the experience orintuition of the researcher. Thus, in order to receive reliable resultsis
necessary to use a combination of the methods mentioned above.
References
1. Deignan, Alice. Metaphor and Corpus Linguistics. Amsterdam: John Benjamins,
2005. 235p.
2. MacEnery, Tony, Corpus Linguistics: AnIntroduction. Edinburgh: Edinburgh
University Press, 2001. 235p.
3. Partington, Alan, In: Anatol Stefanowitsch and Stephan Thomas Gries (eds). Cor-
pus-based Approaches to Metaphor And Metonymy, 267304. Berlin: Walter
de Gruyter, 2006.
4. Goatly, Andrew. The Language of Metaphors. London: Routledge, 1997.
5. Stefanowitsch, Anatol, Corpus-based Approaches to Metaphor and Metonymy.
In: Anatol Stefanowitsch and Stefan Th Gries (eds). Corpus-based approaches to
metaphor and metonymy, 117. Berlin: De Gruyter Mouton, 2006.
172