Task #8: Metonymy Resolution at SemEval-2007


Katja Markert
School of Computing, University of Leeds, UK

Malvina Nissim
University of Bologna, Italy
and Institute for Cognitive Science and Technology, CNR, Italy

Please contact one of the organisers if you are interested in participating or have any questions about the nature of the task.

The Phenomenon

Metonymy is a form of figurative speech, in which one expression is used to refer to the standard referent of a related one. For example, in

(1) he was shocked by Vietnam

"Vietnam", the name of a location, refers to an event (a war) that happened there. Similar in

(2) BMW slipped 4p to 31p

(3) The BMW slowed down

"BMW" ,the name of a company, stands for its index on the stock market or a vehicle manufactured . by BMW respectively.

The importance of resolving metonymies has been shown for a variety of NLP tasks, such as machine translation (Kamei and Wakao, 1992), question answering (Stallard, 1993), and anaphora resolution (Harabagiu, 1998; Markert and Hahn, 2002).

The Task

Although metonymic readings are potentially open ended and can be innovative, most of them tend to follow specific patterns. Many other location names, for instance, can be used in the same fashion as Vietnam in (1) above. Thus, given a semantic class (e.g. location), one can specify several regular metonymic shifts (e.g. place-for-event) that instances of the class are likely to undergo.

The task is a lexical sample task for English. Participants have to automatically classify preselected expressions of a particular semantic class (such as country names) as having a literal or a metonymic reading, given a four-sentence context. If a metonymic reading is selected, a further specification into prespecified metonymic patterns (such as place-for-event or company-for-stock) or, alternatively, recognition as an innovative reading, is necessary. For SemEval-1 the task is restricted to location and company names. Standard accuracy as well as precision, recall and balanced f-score with respect to each reading type will be used to assess performance.

Apart from stimulating dedicated metonymy systems, the formulation of the task as a classification task allows adaptation and extension of existing WSD systems, for which it offers the following new challenges:

First, training and testing is not necessarily done on the same word (e.g. train on 'bank' and test on 'bank'), rather on possibly different instances of the same class (e.g. train on 'France' and test on 'Britain'). Second, metonymies do not often cross topical boundaries, so that co-occurrences- and 'one-sense-per-discourse'-based approaches are unlikely to be helpful. Third, there is a portion of innovative metonymic readings that do not fall into any of the prespecified patterns, so that yet different methods must be developed to handle these.

Participating Systems

Participation of both supervised or unsupervised systems is possible. Partial participations that only make decisions for part of the testing data are also welcome.