to move myself physically closer to the stars.” Rather, von Braun—who was attempting to popularize the concept of space travel with the American public, and thereby win state funding for his expensive scientific work—employed a basic metaphor that allows speakers to treat PURPOSES AS DESTINATIONS. Thus, von Braun saw the achievement of interstellar space travel as his goal, and expressed this goal using the language of physical destinations.
But von Braun, an ex-member of the Nazi party, was also the controversial creator of the German V-2 missiles that rained down on Britain during World War II. (He had been spirited away to the U.S. as part of operation paperclip at the end of the war, when the American and Russian militaries competed to round up the brains of the German rocket program.) The comedian Mort Sahl took full advantage of this fact, and of von Braun’s use of the PURPOSES ARE DESTINATIONS metaphor, to cheekily propose an alternative title for the scientist’s book: “I aim for the stars but I keep hitting London.” Clearly, the generic and highly reusable metaphor PURPOSES ARE DESTINATIONS is of a different character than the specific utterances that are constructed from it. Researchers refer to the former as a conceptual metaphor, for it resides at the level of thinking and ideas, and to the latter as a linguistic metaphor. Explicating the relationship between the former and the latter is one of the goals of contemporary cognitive and computational approaches to metaphor, and so we shall return to this relationship many times in this book. We’ll also meet the conceptual metaphor PURPOSES ARE DESTINATIONS again in a later chapter.
We each use conventional metaphors every day, perhaps without even realizing it, yet we each have the ability to elaborate on these standard-issue constructions in our own way, to inject our own voice and personality into what we say. Consider the following quote from Twitter co-founder Biz Stone, who writes in his autobiography of how he came to be an early employee of Google:
I didn’t know it at the time, but behind the scenes Evan [Williams] had to pull strings in order to hire me [at Google] (Stone [2014]).
The idea of exerting influence on others by pulling [their] strings is deeply entrenched in the English language. We describe a bargain or an offer of special treatment as “no strings attached” when we believe the giver is not seeking to unduly influence us, which is to say “to pull our strings.” An emotional appeal that hits its mark is said to “tug at our heartstrings,” while we might describe a master manipulator as being able to “play someone like a violin.” A mother’s continued influence on an adult child is often given metaphoric form with the phrase “apron strings,” and any effort (by mother or child) to curtail this influence is described as “cutting the apron strings.” Stone’s use of the strings metaphor in the context of the idiom “behind the scenes” might also bring to mind images of the pulleys and ropes with which stagehands lower and raise the curtain in a theatrical production (indeed, the word “scene” gets its meaning from the piece of cloth that was draped behind the stage in ancient theatres). Metaphors are much more than the stuff of fancy wordplay, and we use them to do much more than give our messages an attractive sheen: metaphors engage fully with the mechanisms of thought, allowing us to spark associations, insights, and other metaphors in the mind of a hearer even when we are using the most conventionalized of figures. We can think of these figures as being made of clay; convention has given them their shape, but we can add fine detail of our own, or bend them further to our own meanings. Let’s look at the larger context of Stone’s metaphor:
I didn’t know it at the time, but behind the scenes Evan had to pull strings in order to hire me. Actually, they were more like ropes. Or cables—the kind that hold up suspension bridges (Stone [2014]).
This elaboration should dispel any doubts about whether the pulling [one’s] strings metaphor is nothing more than an arbitrary idiom that speakers learn to repeat in whatever context that suits. Rather, a metaphor—even a highly conventional metaphor—establishes a frame of thought that encourages us to think in a particular way. Once it has our attention and draws us in, we are free to explore it, question it, and customize it as we see fit. For instance, we might ask how pulling on metaphorical strings might influence another person. If this other person is a weak-willed puppet, a lightweight player, or a minor cog in the machine (notice how effortlessly one metaphor leads to others), then not much effort is needed to exert influence, and so its strings will be light and easy to pull. But if this other person is a significant cog—a so-called big wheel or a heavyweight player—greater effort is needed to achieve any influence, requiring strings of greater thickness and tensile strength. Notice how the metaphor encourages us to think in the source domain (the domain of strings, cogs, pulleys, puppets, etc.) and to transfer our insights from here into the target domain (the domain of corporate decision-making). In Stone’s example, his metaphor leads us to believe that his friend Evan needed to perform Herculean efforts on his behalf, to influence some very powerful people at Google by pulling on some very heavy-duty strings. As we’ll see repeatedly throughout this book, even the most innocuous metaphors conceal a wealth of complexity, both in terms of their underlying representations and the cognitive/computational processes that are needed to understand them. This hidden complexity is a large part of what gives metaphors their allure for the computationalist.
However, although metaphor has a long and illustrious history of academic study, in both philosophy and linguistics, it remains a niche area in the computational study of language. For although most Natural Language Processing (NLP) researchers would readily acknowledge the ubiquity of metaphor in language, metaphor is a complex phenomenon and a hard engineering problem that continues, for the most part, to be wait-listed by the NLP community. To the application-minded, there are simply too many other problems of more practical and immediate interest—concerning syntax, semantics, inference, sentiment, co-reference, under-specification, and so on—that jump to the top of the community’s collective to-do list. That metaphor touches on all of these problems and more is often seen as beside the point, although ultimately it is very much to the point: the problem of metaphor is just too big, too unwieldy, and too knowledge-hungry to be tackled first. Better to get a handle on all the other problems first, to obtain an algorithmic understanding of the workings of language that can later be enriched by a computational model of metaphor. This makes good engineering sense, if little philosophical or cognitive sense.
The purpose of this book is to demonstrate that this conventional wisdom is predicated upon a false dichotomy: researchers can build figurative-language processing systems that are practical and efficient and cognitively plausible, and which also reflect an understanding of the profound philosophical issues involved. Indeed, it is difficult for the computationally-minded researcher to explore aspects of metaphor that have not been previously visited by philosophers or psychologists or by earlier computationalists in its long and illustrious history of academic analysis. If there is little in the field of metaphor that one can truly call “virgin territory,” it is nonetheless a field of many interesting landmarks that rewards careful viewing and repeat visits. We have written this book to be a comprehensive guide to the major landmarks in the computational treatment of metaphor and hope the reader will find it a useful map to this fascinating phenomenon’s many attractions.
CHAPTER 2
Computational Approaches to Metaphor: Theoretical Foundations
2.1 THE WHAT, WHY AND HOW OF METAPHOR
Metaphor is both pervasive and evasive: ubiquitous in language, yet remarkably hard to pin down in formal terms. Yet the fact that there exists no single, definitive perspective on metaphor is very much in keeping with its chimerical nature. For metaphor is a highly productive mechanism that allows us to create a panoply of viewpoints on any concept we care to consider, including metaphor itself. Indeed, it scarcely seems possible to say anything meaningful at all about metaphor without first resorting to one kind of metaphor or another. So we talk of tenors and vehicles, sources
