The Stone and the Shell

I’ve asserted several times that flaws in optical character recognition (OCR) are not a crippling problem for the English part of the Google dataset, after 1820. Readers may wonder where I get that confidence, since it’s easy to generate a graph like this for almost any short, random string of letters:

It’s true that the OCR process is imperfect, especially with older typography, and produces some garbage strings of letters. You see a lot of these if you browse Google Books in earlier periods. The researchers who created the ngram viewer did filter out the volumes with the worst OCR. So the quality of OCR here is higher than you’ll see in Google Books at large — but not perfect.

I tried to create “xec” as a nonsense string, but there are surprisingly few strings of complete nonsense. It turns out that “xec” occurs for all kinds of legitimate reasons: it appears in math, as a model number, and as a middle name in India. But the occurrences before 1850 that look like the Chicago skyline are mostly OCR noise. Now, the largest of these is three millionths of a percent (10^-6). By contrast, a moderately uncommon word like “apprehend” ranges from a frequency of two thousandths of a percent (10^-3) in 1700 to about two ten-thousandths of a percent today (10^-4). So we’re looking at a spike that’s about 1% of the minimum frequency of a moderately uncommon word.

In the aggregate, OCR failures like this are going to reduce the frequency of all words in the corpus significantly. So one shouldn’t use the Google dataset to make strong claims about the absolute frequency of any word. But “xec” occurs randomly enough that it’s not going to pose a real problem for relative comparisons between words and periods. Here’s a somewhat more worrying problem:

English unfortunately has a lot of letters that look like little bumps, so “hirn” is a very common OCR error for “him.” Two problems leap out here. First, the scale of the error is larger. At its peak, it’s four ten-thousandths of a percent (10^-4), which is comparable to the frequency of an uncommon word. Second, and more importantly, the error is distributed very unequally; it increases as one goes back in time (because print quality is poorer), which might potentially skew the results of a diachronic graph by reducing the frequency of “him” in the early 18c. But as you can see, this doesn’t happen to any significant degree:

“Hirn” is a very common error because “him” is a very common word, averaging around a quarter of a percent in 1750. The error in this case is about one thousandth the size of the word itself, which is why “hirn” totally disappears on this graph. So even if we postulate that there are twenty equally common ways of getting “him” wrong in the OCR (which I doubt), this is not going to be a crippling problem. It’s a much less significant obstacle than the random year-to-year variability of sampling in the early eighteenth century, caused by a small dataset, which becomes visible here because I’ve set the smoothing to “0” instead of using my usual setting of “5.”

The take-away here is that one needs to be cautious before 1820 for a number of reasons. Bad OCR is the most visible of those reasons, and the one most likely to scandalize people, but (except for the predictable f/s substitution before 1820), it’s actually not as significant a problem as the small size of the dataset itself. Which is why I think the relatively large size of the Google dataset outweighs its imperfections.

By the way, the mean frequency of all words in the lexicon does decline over time, as the size of the lexicon grows, but that subtle shift is probably not the primary explanation for the downward slope of “him.” “Her” increases in frequency from 1700 to the present; “the” remains largely stable. The expansion of the lexicon, and proliferation of nonfiction genres, does however give us a good reason not to over-read slight declines in frequency. A word doesn’t have to be displaced by anything in particular; it can be displaced by everything in the aggregate.

An even better reason not to over-read changes of 5-10% is just that — frankly — no one is going to care about them. The connection between word frequency and discourse content is still very fuzzy; we’re not in a position to assume that all changes are significant. If the ngram viewer were mostly revealing this sort of subtle variation I might be one of the people who dismiss it as trivial bean-counting. In fact, it’s revealing shifts on a much larger scale, that amount to qualitative change: the space allotted to words for color seems to have grown more than threefold between 1700 and 1940, and possibly more than tenfold in fiction.

This is the fundamental reason why I’m not scandalized by OCR errors. We’re looking at a domain where the minimum threshhold for significance is very high from the start, because humanists basically aren’t yet convinced that changes in frequency matter at all. It’s unlikely that we’re going to spend much time arguing about phenomena subtle enough for OCR errors to make a difference.

This isn’t to deny that one has to be cautious. There are real pitfalls in this tool. In the 18c, its case sensitivity and tendency to substitute f for s become huge problems. It also doesn’t know anything about spelling variants (antient/ancient,changed/changd) or morphology (run/ran). And once in a great while you run into something like this:

“Hmm,” I thought. “That’s odd. One doesn’t normally see straight-sided plateaus outside the 18c, where the sample size is small enough to generate spikes. Let’s have a bit of a closer look and turn off smoothing.”

Yep, that’s odd. My initial thought was the overwhelming power of the movie Love Story, but that came out 1970, not 1972.

I’m actually not certain what kind of error this is — if it’s an error at all. (Some crazy-looking early 18c spikes in the names of colors turn out to be Isaac Newton’s Opticks.) But this only appears in the fiction corpus and in the general English corpus; it disappears in American English and British English (which were constructed separately and are not simply subsets of English). Perhaps a short-lived series of romance novels with “romantic” in the running header at the top of every page? But I’ve browsed Google Books for 1972 and haven’t found the culprit yet. Maybe this is an ill-advised Easter egg left by someone who got engaged then.

Now, I have to say that I’ve looked at hundreds and hundreds of ngrams, and this is the only case where I’ve stumbled on something flatly inexplicable. Clearly you have to have your wits about you when you’re working with this dataset; it’s still a construction site. It helps to write “case-sensitive” on the back of your hand, to keep smoothing set relatively low, to check different corpora against each other, to browse examples — and it’s wise to cross-check the whole Google dataset against another archive where possible. But this is the sort of routine skepticism we should always be applying to scholarly hypotheses, whether they’re based on three texts or on three million.

I’ve written several posts now on the way related terms (especially simple physical adjectives) tend to parallel each other in the Google dataset. The names of primary colors rise and fall together. So do “hot” and “cold,” “wet” and “dry,” “thin” and “thick,” “clean” and “dirty,” and the names of the seasons.

These correlations tend to be strongest in the fiction corpus, but most of them hold in other corpora as well. Moreover, all the terms I just mentioned seem to have a minimum value in the early nineteenth century (around 1820) and a maximum around 1940.

Since I’ve listed a lot of binary oppositions, and playfully channeled Lévi-Strauss at the end of an earlier post, it may be time for me to offer a few disclaimers.

The title of the article published in Science was “Quantitative Analysis of Culture Using Millions of Digitized Books.” But I’m afraid I agree with Matthew Jockers, among others, that in this context the word “culture” is unhelpful. To be fair to the “culturomics” team, it’s an unhelpfully vague word in most other contexts too. Writers often invoke “culture” when they need to connect phenomena without an evident causal connection. The New York Times wedding pages may seem to have nothing to do with Facebook. But all I have to do is characterize them as coordinate expressions of a single “culture of narcissism” and — ta da!

Some of the blame for this habit of argument may rest with structural anthropologists who mapped different kinds of behavior onto each other (kinship relations, language, myth), and characterized them as expressions of the same underlying cultural oppositions. So when I start enumerating oppositions, I should stress that I don’t think the Google dataset proves a structuralist theory of culture, or that we have to assume one in order to use it.

I want to suggest that changes in diction are meaningful phenomena in their own right, and that the task of interpreting them is essentially descriptive. We don’t have to read diction as a symptom of something more abstract like culture. Of course, to say that this is a descriptive task is not to deny that it involves interpretation. Patterns don’t foreground themselves.

There’s interpretation involved in pairing “thick” and “thin,” just as there is whenever we highlight a pattern in a literary work. But we’re describing a pattern perceptible in the history of diction, not speculating about a hidden cultural agency.

To explain these patterns causally, they may need to be broken into smaller pieces. It’s possible, for instance, that the commonest concrete adjectives became less frequent in the early nineteenth century because they got partly displaced by Latinate near-synonyms, but became infrequent in the late twentieth century for a completely different reason — say, because adjectives in general became less common in prose. (I’m just speculating here.) Genres will also need to be distinguished. It seems likely that concrete adjectives peak around 1940 partly because modernist novels explore the hot, wet phenomenology of life, and partly because pulpy sci-fi stories describe the hot, wet jungles of Venus.

The relative contributions of different genres will need to be disentangled before we really understand what happened, and Google unfortunately is not going to do much to help us there.

All this is to say that I’m not offering an explanation when I mention structuralism. I certainly don’t mean to invoke “culture” as an explanation for these patterns. It will be far more interesting to understand them, eventually, as consequences of specific generic and stylistic shifts.

I mention structuralism only as a (very loose!) metaphor for one way of extracting literary significance from the history of diction. Right now a lot of humanists have the impression that this sort of interpretation would have to rely on sympathetic magic: the fact that the word “sentimental” peaked around 1930 would only interest us if we could assume that this somehow made the Thirties the most sentimental decade of all time. (Kirstin Wilcox pointed me to the history of “sentimental,” btw.)

Focusing on sets of antonyms has the advantage of ruling out this sort of sympathetic magic. The world can’t have become at once thinner and thicker, wetter and drier, in the early 20th century. When both parts of an opposition change in correlated ways, the explanation required is clearly stylistic. To put this another way, wet/dry and thin/thick are connected not by a mysterious black box called “culture” but by the patterns of selection writers had to learn in order to reproduce a historically specific style.

The dataset that Google made public last week isn’t perfect. As Natalie Binder among others has pointed out, the dataset contains many OCR (optical character recognition) errors, and at least a few errors in dating. (UPDATE 12/22: It is worth noting, however, that the dataset will have many fewer errors than Google Books itself, because the dataset is based on a subset of volumes with relatively clean OCR.)

Moreover, as Dennis Baron argues in The Web of Language, “books don’t always reflect the spoken language accurately.” Informal words like “hello” are likely to be underrepresented in books.

The utility of the dataset is even more importantly reduced by Google’s decision to strip out all information about context of original occurrence, as Mark Liberman has noted. If researchers had unfettered access to the full text of original works, we could draw much more interesting conclusions about context, genre, and authorship.

Finally, I would add that — even with the present structure of the dataset — it’s possible to imagine search strategies other than simply graphing the frequencies of individual words and phrases, one by one. The ngram viewer is an elegant interface, but a limited one.

All true. But the Google dataset is also turning out to be tremendously useful, and it’s likely to become even more useful as researchers refine it and develop more flexible ways to query it.

Of course, it has to be used appropriately. This is not a tool you should use if you want to know exactly how often Laurence Sterne referred to noses. It’s a tool for statistical questions about the written language that involve very large numbers of examples. When it’s applied to questions on that scale, the OCR errors in the English corpus (after 1820) are not significant enough to prevent the ngram viewer from producing useful results. Before 1820 there are more significant OCR problems, especially with the substitution of f for “long s.” But even there, I don’t see the problem as insuperable; there are straightforward ways for researchers to compensate for the most predictable OCR errors.

The larger critique being leveled at the ngram viewer, by Natalie Binder and many other humanists, is that it’s impossible to know what an individual graph measures. Complex words have multiple meanings, Binder reminds us, so how should we interpret a graph showing a decline in the frequency of “nature”? How should we interpret a correlation between the increasing frequency of “vampire” and the declining frequency of “dilettante”?

The saying that correlation doesn’t prove causation definitely needs to be underlined in this domain. There are so many words in the language that a huge number of them will always correlate in largely accidental ways. More generally, it’s true that, in most cases, a graph of word frequency will not by itself tell us very much. You have to have some cultural context before the increasing frequency of “vampire” in the late twentieth century is going to mean anything at all to you. But of course, this is true of all historical evidence: no single poem or novel, in isolation, can tell us what was happening culturally around 1800. You need to compare different texts and authors from different social groups; it may be helpful to know that there was a revolution in France, and so on.

What puzzles me about humanistic disdain for the ngram viewer is that it often seems to presume that a piece of evidence must be legible in itself — naked and free of all context — in order to have any significance at all. If a graph doesn’t have a single determinate meaning, read from its face as easily as the value of a coin, then what is it good for? This critique seems to take hyper-positivism as a premise in order to refute a rather mild and contextual empiricism.

In short, the evidence produced by Google’s new tool is imperfect. It will have to be interpreted sensitively, by people who understand how it was produced. And it will need to be supplemented by multiple kinds of context (literary, social, political), before it acquires much historical significance. But these things are also true of all the other forms of evidence humanists invoke.

It seems likely that humanists are reluctant to take this kind of evidence seriously not because they find it too loose and indeterminate, but because they fear that the superficial certainty of quantitative evidence will seduce people away from more difficult kinds of interpretation. This concern can easily be exaggerated. If an awareness of social history doesn’t prevent us from reading sensitively (and I don’t think it does), then the much weaker evidence provided by text-mining isn’t likely to do so either. I’m reminded of an observation Matt Yglesias made in a different (political) context: that people are in general liable to take “an unduly zero-sum view of human interactions.” Different kinds of evidence needn’t be construed as competitive; they might conceivably enrich each other.