Making and Understanding Affective Metaphors with
Metaphor Magnet is an online software application that allows you to explore the space of affective conceptual metaphors. The Metaphor Magnet application can also be used as a service by other NLP applications, returning XML documents to the client. Details of this XML functionality are provided at the end of this page.
The goal of Metaphor Magnet is to find and exploit commonplace metaphors in everyday texts (in this case, in the Google n-grams) and to use these mappings to interpret more novel metaphors using an expansionary approach.
Using the Metaphor Magnet app, a user inputs a metaphor as in the following:
This is an affective metaphor: the vehicle game is marked with + to indicate that a positive spin is intended – the user is here seeking an uplifting interpretation of the metaphor life is a game (conversely, a more cynical interpretation can be obtained using the input life is a –game).
As shown below, Metaphor Magnet offers its interpretation in two parts: on the right, a term cloud shows positive metaphors for the source concept game; on the left, a term cloud shows these metaphors projected into the domain of life.
In contrast, a negative spin (life is a –game) yields the following clouds:
The most salient metaphors for the source concept game are not always the most salient for the target concept. The qualities that are shown to be salient in aggregate for the source are used to determine the most salient viewpoints on the target. Consider this metaphor: love is a +drug
Note how the descriptions in the left cloud (for the source, +drug) offer mostly literal views on the topic drug. When viewed through the lens of the target concept love, the most salient qualities are healing, satisfying and intoxicating.
In effect, the left cloud shows an expansive interpretation of a particular affective spin on the source concept. The right cloud shows how this expansion (e.g. for positively describing drugs) leads to specific metaphoric views on the target concept. Notice how very different viewpoints are suggested when the user’s query instead becomes: love is a –drug
A user can also pose a figurative query in the form of a simile. This form allows the user to focus the interpretation on a specific quality. For instance, consider the query Google is as –powerful as Microsoft and the interpretation it yields:
The converse, Google is as +powerful as Microsoft, yields this interpretation:
Each element of an interpretation is clickable. For instance, clicking on the element revered:pioneer in the left window brings up a specific interpretation:
In this window, the left cloud shows the linguistic evidence for viewing Google as a pioneer (this evidence is found in the Google n-grams). The right panel shows the results of a Google web search using this evidence as the basis of query expansion.
Metaphor Magnet as a Remote Service: XML output
The above screen shots show Metaphor Magnet operating as a conventional web application, producing HTML output in response to user inputs.
Metaphor Magnet can also be used as a metaphor generation/interpretation service by other NLP applications that wish to exploit its figurative capabilities.
Client applications can interact with Metaphor Magnet by opening a URL connection; the appropriate parameters should be contained in the URL. For instance, here is one such URL:
Notice the parameter &xml=true at the end of this URL: this indicates to Metaphor Magnet that a structured XML document, rather than HTML, should be returned to the client. Below we see XML output for the input life is a +game
Simple experimentation will show client developers the expected form of URLs to be sent to Metaphor Magnet. To view the XML output of any query, the app contains a View XML link on each page:
Two sample XML documents are included with this zip archive:
The latter document contains evidence for the element revered:king in the metaphorical query Google is as +powerful as Microsoft
ReferencesWhen referring to this research/application, please cite one of the following:
- Veale, T. (2012). Exploding the Creativity Myth: The Computational Foundations of Linguistic Creativity. Bloomsbury Academic. [sample chapter]
- Veale, T. (2012). A Context-sensitive, Multi-faceted model of Lexico-Conceptual Affect. In Proc. of ACL’2012, the 50th Annual Conference of the Association for Computational Linguistics, Jeju, South Korea. [pdf]
- Veale, T. and Li, G. (2012). Specifying Viewpoint and Information Need with Affective Metaphors: A System Demonstration of Metaphor Magnet. In Proc. of ACL’2012, the 50th Annual Conference of the Association for Computational Linguistics, Jeju, South Korea. [pdf]