DHQ: Digital Humanities Quarterly
2009
Volume 3 Number 2

# Mining Eighteenth Century Ontologies: Machine Learning and Knowledge Classification in the Encyclopédie

## Abstract

The Encyclopédie of Denis Diderot and Jean le Rond d'Alembert was one of the most important and revolutionary intellectual products of the French Enlightenment. Mobilizing many of the great – and the not-so-great – philosophes of the 18th century, the Encyclopédie was a massive reference work for the arts and sciences, which sought to organize and transmit the totality of human knowledge while at the same time serving as a vehicle for critical thinking. In its digital form, it is a highly structured corpus; some 55,000 of its 77,000 articles were labeled with classes of knowledge by the editors making it a perfect sandbox for experiments with supervised learning algorithms. In this study, we train a Naive Bayesian classifier on the labeled articles and use this model to determine class membership for the remaining articles. This model is then used to make binary comparisons between labeled texts from different classes in an effort to extract the most important features in terms of class distinction. Re-applying the model onto the original classified articles leads us to question our previous assumptions about the consistency and coherency of the ontology developed by the Encyclopedists. Finally, by applying this model to another corpus from 18th century France, the Journal de Trévoux, or Mémoires pour l'Histoire des Sciences & des Beaux-Arts, new light is shed on the domain of Literature as it was understood and defined by 18th century writers.

# Introduction

One of the crowning achievements of the 18th-century Enlightenment was the Encyclopédie ou Dictionnaire raisonné des sciences, des arts et des métiers, par une société de gens de lettres, edited by Diderot and d'Alembert. Published in Paris between 1751 and 1772, in 17 volumes of text and 11 volumes of plates, it contains some 77,000 articles written by more than 140 contributors. The Encyclopédie was a massive reference work for the arts and sciences, as well as a machine de guerre that served to propagate Enlightenment ideas. The impact of the Encyclopédie was enormous. Through its attempt to classify learning and to open all domains of human activity to its readers, the Encyclopédie gave expression to many of the most important intellectual and social developments of its time.[1]
The scale and ambition of the Encyclopédie inspired its editors to adopt three distinct modes of organization which, taken together, Diderot described as encyclopedic: dictionary, hierarchical classification, and the renvois (cross-references). The interaction of these three modes has led modern commentators to describe the Encyclopédie as an “ancestor of hypertext” and to depict Diderot as “ l'internaute d'hier ” (“websurfer avant la lettre ”) [Brian 1998]. D'Alembert underscores the importance of the organization of knowledge in the Discours Préliminaire:

As an Encyclopedia, it is to set forth the order and connection of the parts of human knowledge. As a Reasoned Dictionary of the Sciences, Arts, and Trades, it is to contain the general principles that form the basis of each science and each art...and the most essential facts that make up the body and substance of each.[2]

Of the three modes of organization, the dictionary mode (organization of entries in alphabetical order) is certainly the simplest and the most arbitrary. The second mode of organization is classification, wherein each dictionary entry is assigned to a “class of knowledge,” placing it within the “order” of human understanding, as depicted in the Système Figuré des connaissances humaines. Modeled after Bacon's classification of knowledge and Enlightenment theories of epistemology, all understanding is founded upon memory, reason, or imagination, with numerous categories and sub-categories branching out from these three faculties.[3] However, simply placing an entry into this hierarchy of knowledge was insufficient to indicate the interconnections of knowledge. Thus, Diderot created an extensive system of renvois , or cross-references, the third mode of organization, providing a lattice of interconnections between individual leaves of the tree as well as between classes of knowledge.[4]
The central role of the classification system in the intellectual objectives of the Encyclopédie editors is indicated by the extent to which it has been discussed and debated by both contemporary scholars and later researchers. The editors were remarkably diligent in assigning classes of knowledge to each article and sub-article. Of the 77,085 main and sub articles, we have identified 55,248 as having classes of knowledge.[5] The editors were, however, somewhat less diligent in maintaining a precisely controlled list. Thus the classifications as found in the text are an amalgam of abbreviations, conflations, and singular categories that are not found on the Système Figuré at all. We have recently completed orthographic normalization of the classes of knowledge assigned to each article,[6] resulting in some 54,289 articles with 2,620 normalized classes of knowledge. The twenty most frequent classifications by number of articles are:
 Géographie 5513 Géographie moderne 4794 Géographie ancienne 3084 Jurisprudence 2396 Grammaire 2304 Marine 1894 Commerce 1483 Histoire naturelle. Botanique 1277 Histoire moderne 1194 Mythologie 1115 Histoire naturelle 1069 Histoire ancienne 889 Medecine 796 Architecture 730 Jardinage 689 Littérature 682 Maréchallerie 627 Botanique 614 Histoire ecclésiastique 614 Théologie 517
Table 1.
Original counts of top 20 most frequent classifications by number of articles
Like the Système Figuré, these classifications are a reflection of how knowledge was ordered and classified in the 18th century. This paper reports the results of experiments using machine learning and data mining techniques to understand and exploit this unique resource.

# Preliminary Objectives

Our initial experimental objectives were threefold:
1. To train our classifier on all of the normalized classified articles of the Encyclopédie and then to apply the resulting model to the 22,796 unclassified articles, labeling each with a predicted class of knowledge.
2. Using the same model, we then sought to reclassify all of the classified articles, verifying the relative success of our classifier by examining the misclassified instances and then exploring the relationships of discrete classes of knowledge to one another through feature set analysis.
3. Finally, we wanted to apply the classification model generated from the Encyclopédie to other 18th century texts, allowing us to evaluate the applicability of this ontology to sources outside of the encyclopedic enterprise.

# Methodology

Before we could begin any classification tasks we needed to preprocess the data into a usable format. To this end, we extracted the text from all articles and sub-articles in the Encyclopédie with a normalized class of knowledge. Explicit markers of class membership, present at the beginnings of these articles (e.g., Comm. for Commerce, Phys. for Physics, etc.) were removed to ensure that they would not provide facile criteria for classification. The texts were then tokenized and lemmatized[7] automatically, and frequencies of words and lemmas were computed both globally and for each article. Words and lemmas were then used as attributes, and vectors for each article were generated from the number of occurrences of each attribute in that article.
Text categorization is an extensively studied subfield of information science and accordingly, there exist many time-tested classification algorithms, each with their own strengths and weaknesses. We chose to implement a Multinomial Naive Bayesian classifier because of its simplicity and efficacy on large corpora.[8] Multinomial Naive Bayes (MNB) treats documents as bags of words, where word order is considered irrelevant. Essentially, we measure the likelihood of words occurring in a given class by calculating how many times they occur in all documents with known classes. These conditional probabilities allow us to guess the most “probable” class of an unclassified article based on the frequency of the words that occur within it. MNB assumes that the probability of a word occurring within a document is independent of the words that occur around it, an assumption that we know to be false. Yet, despite this, MNB is known to perform very well on large data sets and in fact, has been shown to produce more accurate results than many other classification algorithms.[9]
We designed a test bed using this implementation that allowed for easy modification of several parameters including the minimum word count required for an article to be included in our model, as well as the minimum and maximum number of occurrences across the entire corpus required for a word or lemma to be included in our vocabulary. Massaging the data in this way, we can avoid, as much as possible, skewed results caused by high frequency function words or highly weighted words occurring in a relatively small number of articles. After several rounds of experimentation we found that our classifier was most successful when excluding articles of 25 words or fewer and words (lemmas) occurring in less than 4 articles.[10]

# Results

## Classifying the Unclassified

Our first classification task was to train the classifier on the 54,289 articles that were assigned categories of knowledge by the editors of the Encyclopédie and to then apply this model to the 22,796 unclassified articles in an attempt to predict class membership for the articles in question. Once classified, the twenty most frequent labels for the newly classified articles were:
 Jurisprudence 4276 Art méchanique 1260 Géographie 828 Commerce 802 Anatomie 643 Marine 557 Histoire moderne 475 Architecture 435 Grammaire 402 Histoire naturelle. Ornithologie 367 Medecine 363 Géographie moderne 347 Art militaire 311 Histoire ecclésiastique 308 Géométrie 306 Géographie ancienne 306 Musique 295 Astronomie 274 Histoire naturelle. Botanique 266 Théologie 215
Table 2.
Counts of top 20 most frequent classifications for previously unclassified articles.
While this distribution of classes bears some resemblance to the overall distribution cited above, we have no real way of verifying the accuracy of the classifier given the unknown content/classes of the unclassified articles.[11] While it is entirely plausible that 19% of the unclassified articles are concerned with Jurisprudence in a general sense, it is also likely that Jurisprudence (which represents only 5% of the classified articles) becomes of sort of “catch-all” category for the classifier into which articles concerned with any specific aspect of law (i.e. droit romain, droit canonique, droit civil, etc.) are grouped.
The sample of results we examined reveal that the classifier performed reasonably well. By this, we mean that some classifications seemed right on; some made a good degree of sense, while others were perhaps a bit too general, failing to accurately represent the specificity of the subject matter. Naturally, the more than 22,000 newly generated classifications could not all be verified by hand, and so we focused on major articles and a selection of smaller ones. We were particularly encouraged by the assigned classifications for the 10 longest unclassified articles. The Discours Préliminaire, d'Alembert's famous preface to the Encyclopédie detailing the intellectual underpinnings of the enterprise, comes back as belonging to the class Philosophy.[12] Going down the list we see that the article “Anatomie” is assigned its own classification in Anatomy and “Chimie” is rightly placed into Chemistry, results we had originally hoped were easy enough for the algorithm to attain. Most of the classifications, however, don't fall into such clear categories. For example, “Venerie” — the art of hunting — was assigned to Natural History; the philosophical article “Eclectisme” to the History of Philosophy, etc. Indeed, while these and the better part of the predicted classifications can be justified on a general level, we had to concede that the overall utility of this task was somewhat questionable. Quite simply, as we mention above, there was just too much data to sort through. The new labels were often interesting, but we were not able to study them easily or thoroughly enough to come to any deeper understanding about how the philosophes structured knowledge or indeed how the Encyclopédie itself fits together. Thus, trying to develop an experiment that could generate more legible results, we decided instead to leverage the information given us by the editors in exploring the known classifications and their relationship to each other and then later, to consider the classification scheme as a whole by examining the general distribution of classes over the entire work as opposed to individual instances.

## Classified vs. Classified — Feature set evaluation

Having run a set of predictive classification experiments on the unclassified articles, our next task was an attempt at what we have named "comparative" classification, wherein we train the classifier on two particular classes, and then reclassify them in an effort to determine how separable they are and to ascertain which features best distinguish articles from the two sets. The goal here has less to do with the accuracy of the classifications than with the feature sets that are generated during the classification task. Any two classes can be compared. Feature weights are generated using the Naive Bayes Perl module written by Ken Williams. These weights represent the conditional probability of a feature f given a class c and are generated based on their frequency in the known articles.[13] To give an example of this sort of comparative classification, 889 articles classified as “Histoire ancienne” were compared with 1194 articles in the “Histoire moderne” class and the following features and their weights were extracted as the most relevant in determining class membership:
 étoient 0.04356 avoit 0.03705 romains 0.02472 avoyer 0.02455 an 0.02341 peuple 0.02271 chez 0.02188 sous 0.02182 tems 0.0217 mot 0.02146 empereur 0.02146 g 0.02122 appelloit 0.0188 premier 0.01847 nous 0.01802 encore 0.01783 après 0.01763 homme 0.01756 dieu 0.01683 rome 0.01683
Table 3.
Conditional probabilities of top 20 most common words in articles from Histoire ancienne
 roi 0.03744 ordre 0.0287 prince 0.0239 sous 0.02314 nommer 0.0218 titre 0.02165 empire 0.02065 chevalier 0.02046 officier 0.02039 tems 0.0203 étoient 0.02003 premier 0.01959 empereur 0.01903 porter 0.01891 état 0.01871 mot 0.01851 avoit 0.01823 prendre 0.01814 maître 0.01814 sans 0.0178
Table 4.
Conditional probabilities of top 20 most common words in articles from Histoire moderne
The overall performance of the classifier came in at 95.63%, which tells us that while both belong to the same “branch” of science, namely History, the ancient and the modern are nonetheless significantly distinguishable from one another. When considering the two lists of features, one immediately notices that the results make good sense, i.e., we find more verbs in the past tense ( avoient , étoient, etc.) in the Ancient History articles as well as references to antiquity ( romains , empereur, rome, etc.). The single-letter feature “g” is the signature of the Abbé Mallet who was the author of a significant portion of the articles on Ancient History. Interestingly, some of the features occur in both lists, which is more than likely a result of the close relationship and dialogue between the two classes. In the feature set for Modern History, we find references to forms of government that quite rightly belong to the modern period ( état, prince, roi, chevalier, etc.) and the notable absence of the word “ dieu ” (“god”).
We can also run comparative classification tasks on seemingly dissimilar classes of knowledge such as the 682 Literature articles and the 200 articles dealing with Physics. Not surprisingly, we obtain a very high rate of success for this sort of classification, in this case, 99.29%:
 nous 0.0333 mot 0.02902 avoit 0.02607 étoient 0.02294 livre 0.02283 ancien 0.02149 je 0.02117 tems 0.0203 bien 0.01997 encore 0.01963 sans 0.01956 vers 0.01907 dieu 0.01835 auteur 0.0182 latin 0.01785 usage 0.01742 devoir 0.01738 notre 0.01695 homme 0.01691 chose 0.01671
Table 5.
Conditional probabilities of top 20 most common words in articles from Littérature.
 corps 0.05394 air 0.04216 eau 0.04185 nous 0.03507 froid 0.02597 chaleur 0.02584 monsieur 0.0258 degré 0.02296 moins 0.02251 fort 0.02242 glace 0.02196 couleur 0.02169 feu 0.0216 lorsque 0.02109 effet 0.02048 peu 0.01991 rayon 0.01986 fluide 0.01962 mouvement 0.01952 trouver 0.01948
Table 6.
Conditional probabilities of top 20 most common words in articles from Physique
The feature scores from this model are what we would expect. The Literature class' most significant features are those words pertaining to language and grammar ( mot, livre, vers, auteur, latin, usage, etc.) whereas the Physics class is dominated by materialistic terminology ( corps, air, eau, degré, fluide, mouvement, etc.) consistent with the scientific writings of the period.
Evaluation of these feature sets can be invaluable when testing certain hypotheses, such as word usage differences across similar disciplines or between authors. In this particular case, the features provide an intuitive illustration of the differences between these two distinct classes of knowledge. While features are simply terms that the algorithm finds statistically representative of a particular class, the feature sets as a whole can also give a snapshot of the make-up of the individual classes or indeed of larger concepts more generally. From the list for literature, for example, we get a sense of the importance that classical Roman authors still had for the philosophes with the terms “ancien” (“ancient”) and latin. “Vers” (“verse”) perhaps reflects the fact that versification was a predominant aspect of literary style, whether in poetic, dramatic, and other writing at that time. Terms such as “mot” (“word”) and “usage” (“use”) might point to the 18th century's expansive, belle-lettristic sense of literature that we discuss below. In a more general manner, feature sets provide us with an expanded thesaurus for any given classification task -- leads for further investigation and study -- that can then be exploited by a more traditional full text analysis system.[14]

## Reclassifying the Classified — the Ontology of the Encyclopédie

Finally, we applied the model assembled for our first experiment — trained on all of the known classifications — onto all of the already classified articles. By this, we mean that we effectively ignored any given classes of knowledge, treating each article as if it were unclassified, and then assigned class membership using the algorithm described above. Here our goal in the results analysis was twofold: first, we were curious as to the overall performance of our classification algorithm, i.e., how well it correctly labeled the known articles; and secondly, we wanted to use these new classifications to examine the outliers or misclassified articles in an attempt to understand better the presumed coherency and consistency of the editors' original classification scheme.
We achieved a 71.4% success rate in the re-categorization of the 54,289 classified articles, a performance that could perhaps be improved with a more accurate morphological stemmer and the inclusion of n-grams as features, fucnctions we intend to implement in the future. Nonetheless, developing a model to reliably guess an article's given class of knowledge is ultimately not our primary concern as even a perfect model, while impressive in terms of performance, could only yield that which we already know, namely the assigned classes of knowledge. The sheer size and complexity of the Encyclopédie, drawing its contents from hundreds of distinct writers, all but guarantees a lower rate of performance for any classification algorithm. This fact need not be discouraging however, as we are more interested in exploring the use of these text mining techniques as knowledge discovery tools, uncovering previously unnoticed connections and classifications, such as the particular use of the class “Literature” outlined below, rather than simply using these approaches as a statistical platform for hypothesis testing.
The twenty most frequent classes after re-classification:
 Géographie 3926 Géographie ancienne 3492 Géographie moderne 3273 Jurisprudence 2552 Commerce 2104 Art méchanique 1662 Histoire naturelle. Botanique 1615 Marine 1575 Histoire moderne 1514 Mythologie 1334 Architecture 1213 Grammaire 1111 Histoire ancienne 1061 Histoire ecclésiastique 781 Medecine 746 Histoire naturelle 727 Littérature 646 Maréchallerie 592 Morale 573 Jardinage 566
Table 7.
The 20 most frequent classes of knowledge by number of articles after re-classification.
When comparing the results to the original classifications we note that the class “Grammar” falls out of the top ten while “Art méchanique,” which is not included in the original top twenty, ranks as the sixth most frequent class. The Grammar class is known to be problematic as Diderot frequently used this seemingly innocuous label to hide more polemical entries.[15] As for the “Art méchanique” category, we suspect that many of the overly specific classes dealing with the mechanical arts were subsumed into this larger, more inclusive set. By and large the rest of the classes are consistent with the overall distribution in the Encyclopédie although the rankings differ slightly.
The most interesting results here come from the examination of misclassified articles, which belie vocabularies that do not belong probabilistically to their assigned categories. Upon analyzing a random subset of the misclassified articles, we identified three distinct types of misclassifications. First, there are articles whose original classification was too infrequent; for example, the article “Accrues” (metal rings used to knit together fishing net) is the sole member of the class “Marchands de Filets” (net merchants) and was placed into the more general class of “Pêche” (fishing). There are also articles whose vocabularies mislead the classifier. One such case is the article “Achées” (a type of worm used in bait-fishing), originally classified as “Pêche,” it was later assigned to the class “Jardinage” (gardening). The article is in fact less a description of anything to do with fishing, but rather contains instructions on how to find and cultivate bait worms in a garden. Finally, there are entries whose predicted class, while incorrect, seems more logical than the original. The article “Tepidarium,” which describes an ancient Roman bathhouse, would appear to have more in common with its predicted class, Architecture, than the one assigned by the editors, namely Literature. Certainly our judgment that the predicted class is more appropriate than the original class of knowledge is biased by our modern epistemological paradigm, but this does not necessarily mean that the original system of classification was entirely consistent and coherent. Naturally then, applying our model onto other 18th century French texts should provide further insight into the power of the classifier and more importantly, into the ontology originally laid out by the philosophes.

## Classification outside of the Encyclopédie

The Journal de Trévoux, or Mémoires pour l'Histoire des Sciences & des Beaux-Arts, was one of the most influential 18th century French periodicals. A sort of literary/scholarly journal reviewing and commenting on a wide variety of contemporary publications, the Journal de Trévoux dealt with almost every discipline of knowledge. Given the great variety of subject matter contained in this collection we felt it would be a natural choice for us to begin studying the relationship of the Encyclopédie ontology to other 18th century texts. Of course, the 18th century in France was a time of intellectual ferment and, as in most historical moments, there existed more than one approach to the classification of the known world. And, indeed, the philosophes' intellectual and political bent made their way of organizing ideas different from that of other thinkers, particularly the Jesuits who were behind the Journal de Trévoux. Knowing this, we wanted to test the degree of overlap between the structure of knowledge in the Encyclopédie and the Journal de Trévoux, discovering in the process the commonalities, differences, and unique aspects of each. We hoped this experiment would give us a "slice of life" look at the intellectual milieu of the day, or at least provide an insight into the presumed differences in discourse between the two camps. To this end, we processed the ARTFL Project's 109 volumes of the Journal de Trévoux, splitting them into 1,027 separate articles. Extending from 1751 to 1757, this collection covers the years during which the editors of the Journal engaged in a fierce polemic with the encyclopédistes concerning the publication of the Encyclopédie.[16] Our previous model, trained on all of the classified Encyclopédie articles, was thus applied to the Trévoux articles, assigning each with a predicted class of knowledge. The twenty most frequent assigned classes are listed below:
 Littérature 317 Morale 86 Géographie moderne 61 Théologie 54 Philosophie 50 Histoire moderne 46 Belles lettres 45 Astronomie 35 Métaphysique 30 Histoire ecclésiastique 26 Physique 22 Art militaire 18 Economie politique 18 Géographie 16 Medecine 16 Histoire romaine 14 Peinture 14 Histoire 14 Histoire naturelle 13 Chimie 12
Table 8.
The 20 most frequent classes of knowledge assigned to the Trévoux articles using the Encyclopédie model.
A cursory glance at these results gives us a general idea about the most significant themes found in the Journal de Trévoux; themes that correspond nicely to our preconceived notions concerning the Journal, its writers and subject matter. Along these lines, it is not surprising to find in a Jesuit publication such as this a greater emphasis on articles about Literature, Morality, Theology, and Philosophy. However, the surprising fact that more than 1/3 of the 1000 articles were assigned the label of Littérature caused us to question somewhat the performance of the classifier and ultimately, to reconsider our modern notion of Literature when applied to the specific instances of this classification.
In the first edition (1694) of the Dictionnaire de l'Académie française the entry for “Littérature” reads thus: “ Litterature. s. f. Erudition, doctrine. Grande litterature. profonde litterature. il est homme de grande litterature. il n'a point de litterature. il a beaucoup de litterature. ” and indeed the definition changes little by the fourth edition of 1762: “ LITTÉRATURE. s.f. Érudition, doctrine. Grande littérature. Profonde littérature. Il est homme de grande littérature. N'avoir point de littérature. Avoir beaucoup de littérature. Un ouvrage plein de littérature. Ce mot regarde proprement les Belles-Lettres. ” The addition of the last sentence, “This word is properly used in regard to Belles Lettres,” in the 1762 edition seemingly restricts this particular form of erudition to the more traditionally literary realm of the “Belles-Lettres,” or Poetry and Rhetoric. The definition offered by the Jesuit editors of the Dictionnaire de Trévoux (1742) differs only slightly from that of the Academy: “LITTÉRATURE, s. f. Doctrine, connoissance profonde des Lettres. Doctrina, litteratura, eruditio.” While these definitions shed little light as to why the Encyclopédie literature class should be so prevalent in the classification of such a diverse collection of articles, many of which deal with the Sciences and Natural History, the ambiguity of this erudite possession of “ littérature ” and “ lettres ” should nonetheless cause us to broaden our understanding of these terms as they were used in the mid-18th century.
We thus began a more thorough investigation of the Literature category by examining five randomly selected articles belonging to the assigned class “Littérature” in the Journal de Trévoux. While some categorizations make sense as literature — e.g., the article “Nouvelles Litteraires,” a sort of literary “news of the day”; and, less convincingly, a commentary on Rousseau's first discourse — others have ostensibly nothing to do with our modern idea of Literature — e.g., articles commenting on a history of jurisprudence, a treatise on diseases, and a compilation of treatises on Physics and Natural History[17]. This apparent anomaly necessarily leads us back to the Encyclopédie and the articles belonging to the class of knowledge Literature, which serve as the basis for these class assignments.
As we mentioned above concerning the reclassification of the article “Tepidarium,” there are a great many articles whose original classifications seem inappropriate. This phenomenon is all the more evident when examining the reclassification of the Literature articles, the majority of which deal more with Ancient History, Mythology, and Architecture than with accepted literary issues. Of the 682 Literature articles, 460 were written by the Chevalier de Jaucourt, author of more than 17,000 Encyclopédie entries. Jaucourt is known to have borrowed extensively from other sources and thus, we attributed these inconsistencies to intellectual laziness, given the enormous number of articles for which he was responsible.[18] Upon closer examination of the Literature class of knowledge however, this characterization proves unjust.
Indeed, the article titled “Littérature” belongs not to its own class of knowledge, but rather to three seemingly unrelated and disparate classes: Sciences, Belles-Lettres, and Antiquity. The text of the article, written by Jaucourt, is in fact a polemic advocating a universal erudition and an expanded definition of what it means to possess a great literature — in a word to be literate. Jaucourt includes a renvoi to the article “Lettres” in an effort to define better this notion of Literature. Following the cross-reference we find that the article in question, “Lettres,” an article that normally falls innocuously amongst numerous similarly titled entries, is the sole member of the class “Encyclopédie,” suggesting that the idea of literacy is essential to the entire encyclopedic enterprise. Here, Jaucourt's understanding of “Lettres” as a much larger category of knowledge than “belles-lettres” or even the Humanities as a whole (les lettres humaines), harkens back to the Classical acceptation of the term which encompassed all areas of human understanding from Epic Poetry to Physics. The inter-connectedness of knowledge, both literary and scientific, is thus the essence of Jaucourt's idea of encyclopedic literacy, wherein “ il en résulte que les lettres & les sciences proprement dites, ont entr'elles l'enchaînement, les liaisons, & les rapports les plus étroits; c'est dans l'Encyclopédie qu'il importe de le démontrer ” (“the result is that Letters and Sciences, properly speaking, have between each other a strong and direct network of links and relationships; it is in the Encyclopaedia that the demonstration of this network becomes important”).[19]

# Conclusions and Future Work

This discovery — that for the writers of the Encyclopédie, Literature as a branch of human understanding included not only what we today consider Literature (les Belles-Lettres) but also Natural History, the Natural Sciences, the study of Antiquity, etc. — was made possible through the machine learning techniques outlined in this paper. We now understand precisely why the classifier, using the ontology of the Encyclopédie, labels so many Trévoux entries as “Littérature.” Furthermore, the presence of this more inclusive view of lettered erudition should come as no surprise given that, as is expressed in the very title of the Encyclopédie, this Dictionnaire raisonné was the work of a society of “Gens de Lettres.”
It would seem, however, that this notion of Literature as a sort of universal erudition did not survive the turmoil of the late 18th century, and by 1798 the fifth edition of the Dictionnaire de l'Académie française had already begun restricting Literature to a purely literary domain, defined as the “ Connoissance des ouvrages, des matières, des règles, et des exemples littéraires ” (our emphasis).[20] 19th century lexicography would move further in this direction, eliminating any mention of “doctrine” or “erudition” in its definition of Literature as “ La science qui comprend la grammaire, l'éloquence et la poésie, et qu'on appelle autrement Belles-lettres ” (Dictionnaire de l'Académie française, 6th edition 1832). By 1872, perhaps as a reflection of the disciplinary codification which took place during the first part of the 19th century, the positivist lexicographer Émile Littré simply defines Literature as the “ Connaissance des belles-lettres ” which is by and large its accepted meaning today.
These modest conclusions lend further weight to our initial view that traditional humanistic inquiry can be enhanced and broadened through the judicious application of machine learning and data mining techniques. As large-scale textual resources such as the Encyclopédie become more readily available to scholars in a digitized format, new search and analysis tools will be needed. It is our opinion that approaches similar to those outlined above can successfully leverage the power of data mining tools for use in the Humanities. And while these techniques can certainly aid in a variety of hypothesis testing and classification tasks, it is our hope that they will also lead to the discovery of new knowledge through the uncovering of previously unnoticed textual connections.
Moving forward, we plan to continue improving the performance of our classifiers through better morphological text extraction, allowing for a greater freedom in the selection of features. Possible features would include n-grams (bi- and tri-grams of surface forms, bi- and tri-grams of lemmas) as well as part of speech information. We are also planning to investigate several different unsupervised machine learning techniques such as vector space analysis, latent semantic indexing (LSI), and several other clustering models. When applied to the Encyclopédie and other 18th century works, these tools will propose connections based on a measure of lexical similarity between arbitrary chunks of texts, whether paragraphs, articles, chapters, or entire works. It is our hope that within this new system, researchers will be able to explore and, more importantly, to evaluate the proposed connections between these articles and texts. Although the connections will be brought to light with the help of computers, it will nonetheless be necessary for scholars to provide the system with the critical element of human scrutiny that is essential to Humanities research.

## Notes

[1] For an exhaustive treatment of the Encyclopédie and its authors, see [Schwab et al. 1971-1984]. For a more general discussion of the work, see [Proust 1995]. The ARTFL implementation of the Encyclopédie is discussed in [Morrissey et al. 2001], and [Andreev et al. 1999].
[2] English translation cited in [Hoyt and Cassirer 1965, xxiii] (our emphasis).
[3] For various representations of the Système Figuré and the Editors' descriptions, see http://www.lib.uchicago.edu/efts/ARTFL/projects/encyc/texts/ and http://artfl.uchicago.edu/cactus/.
[4] Blanchard and Olsen examined the structure of the renvois generating a “mappemonde” of the cross-references and node level classes of knowledge. See [Blanchard and Olsen 2002].
[5] Classes of knowledge were originally extracted automatically using a simple rule-based identifier, written in Perl, based on typographic conventions in the text. A small number of classified articles remain either misclassified or unclassified altogether.
[6] The normalized classifications are the result of a collaborative project with Professor Dena Goodman, Kevin Hawkins, and Benjamin Heller at the University of Michigan.
[7] Lemmatization was accomplished using TreeTagger, a probabilistic part-of-speech tagger and lemmatization utility developed by Helmudt Schmidt under the auspices of the TC Project at the University of Stuttgart. It is freely available for download from the project's official homepage at http://www.ims.uni-stuttgart.de/projekte/corplex/TreeTagger/.
[8] See [Hand and Yu 2001].
[9] See [Witten and Frank 2005, 94–96].
[10] Success can be gauged in various ways. The accuracy achieved when testing our classifier against the same set of classified articles on which it was trained is generally indicative of the Naive Bayesian algorithm's power. However, this sort of validation often leads to over-fitting. Given that our accuracy never exceeded 75%, we can only assume that there are enough anomalies in the data to have compensated for any over-fitting. Additionally, we believe that it is these anomalies that often end up being the most revelatory.
[11] The accuracy of a classifier refers to its ability to correctly predict the label of an unseen instance. Typically, classifier accuracy is estimated using a test set that is independent of the training set, one that is often generated by hand-labeling a set of unseen documents and measuring how often the result of the classifier corresponds to the class given by the human tagger. However, asking modern scholars to classify a large number of articles by hand would necessarily involve not only a large amount of effort, but also a fair amount of guesswork, particularly given the fact that we are dealing with an ontology that is more than 200 years old.
[12] This classification coincides nicely with the conclusions drawn by Martine Groult, one of our collaborators at the CNRS, whose work on the philosophy of Discours préliminaire is summarized here: http://encyclopedie.uchicago.edu/?q=node/162.
[13] It is worth noting that Williams' implementation of the Naive Bayesian classifier uses the log-likelihood ratio of features fi given each class c in question. It also uses a smoothing term to account for probabilities of zero. Because these probabilities are in general very small, their logarithmic weights are very negative. We find it easier to deal with positive numbers, so for display purposes, we make each of these scores positive by raising 2 to the power of that weight. $$classify\left( f_{1},\ldots,f_{n} \right) = {\arg\max}_{c}\text{\ p}(C = c)\prod_{i = 1}^{n}{p\left( F_{i} = f_{i} \middle| C = c \right)} = \log\left( p(c) \right) + \sum_{i = 1}^{n}{\log\left( p\left( f_{i} \middle| c \right) \right)}$$ With smoothing term: $$log\left( p\left( f_{i} \middle| C \right) \right) = \log\left( \left| f_{i} \right| + 1 \right) - \log\left( \left| F_{C} \right| + |F| \right)$$ Williams' Perl module can be downloaded at http://search.cpan.org/~kwilliams/Algorithm-NaiveBayes-0.04/lib/Algorithm/NaiveBayes.pm.
[14] The importance of word usage and the evolution of language in the Encyclopédie, key concepts when considering feature set analysis, are discussed in [Anderson 1984].
[15] The use and abuse of the category “Grammar” by Diderot is one of the subjects treated in [Leca-Tsiomis 1999].
[16] The ARTFL database of the Journal de Trévoux can be found at http://www.lib.uchicago.edu/efts/ARTFL/projects/trevoux.
[17] The full titles of the “Literature” articles we uncovered are as follows: 1) ARTICLE XII. NOUVELLES LITTERAIRES. 2) ARTICLE XXIX. DISCOURS QUI A REMPORTE' le prix à l'Académie de Dijon en l'année 1750, sur cette question proposée par la même Académie: Si le rétablissement des Sciences & des Arts a contribué à épurer les moeurs. Par un Citoyen de Genève. 3) ARTICLE XXXV. HISTOIRE DE LA JURISPRUDENCE Romaine. 4) ARTICLE LXI. TRAITÉ DES MALADIES qu'il est dangereux de guérir. 5) ARTICLE XIV. RECUEIL DE DIFFERENS Traités de Physique & d'Histoire Naturelle.
[18] For a thorough discussion of Jaucourt's contributions to the Encyclopédie, see [Lough 1973].
[19] From the article “Lettres” in the Encyclopédie.
[20] Note that we are focusing exclusively on the definitions of Literature as a discipline, rather than a collection of literary and/or other documents (as in “I have consulted all the available literature on cancer and have found nothing”). This modern usage of literature is expressed somewhat in the dictionaries we mention, i.e. “ L'ensemble des productions littéraires d'une nation, d'un pays, d'une époque. La littérature française. La littérature du moyen âge ” (Émile Littré, Dictionnaire de la langue française (1872-77).

## Works Cited

Anderson 1984  Anderson, Wilda. Between the Library and the Laboratory: The Language of Chemistry in Eighteenth-Century France. Baltimore: Johns Hopkins, 1984.
Andreev et al. 1999  Andreev, Leonid, Jack Iverson, and Mark Olsen. “Re-engineering a War Machine: ARTFL's Encyclopédie ” in Literary and Linguistic Computing 14:1 (1999), 11-28.
Blanchard and Olsen 2002  Blanchard, Gilles, and Mark Olsen. “Le système de renvois dans l'Encyclopédie: une cartographie de la structure des connaissances au XVIIIème siècle” in Recherches sur Diderot et sur l'Encyclopédie, 31-32, April 2002, 45-70.
Brian 1998  Brian, Eric. “L'ancêtre de l'hypertexte,” in Les Cahiers de Science et Vie, 47 (October 1998), 28-38.
Hand and Yu 2001  Hand, David J., and Keming Yu. “Idiot's Bayes — Not So Stupid after All?”, International Statistical Review 69:3 (December 2001), 385-398.
Hoyt and Cassirer 1965  Hoyt, Nelly, and Thomas Cassirer. “Introduction” to Encyclopedia: Selections by Diderot, D'Alembert, and a Society of Men of Letters. Indianapolis: Bobbs-Merrill, 1965.
Leca-Tsiomis 1999  Leca-Tsiomis, Marie. Ecrire l'Encyclopédie: Diderot: de l'usage des dictionnaires à la grammaire philosophique. Oxford: Voltaire Foundation, 1999.
Lough 1973  Lough, John. The Contributors to the “Encyclopédie” . London: Grant and Cutler, 1973.
Morrissey et al. 2001  Morrissey, Robert, Jack Iverson, and Mark Olsen. “Présentation: L'Encyclopédie Electronique” in Robert Morrissey and Philippe Roger, eds., L'Encyclopédie du réseau au livre et du livre au réseau. Paris: Champion, 2001. 17-27.
Proust 1995  Proust, Jacques. Diderot et l'Encyclopédie. Paris: Albin Michel, 1995.
Schwab et al. 1971-1984  Schwab, Richard, Walter Rex, and John Lough. Inventory of Diderot's Encyclopédie, 7 vols. Oxford: Studies on Voltaire and the Eighteenth Century, 1971-1984.
Witten and Frank 2005  Witten, Ian, and Eibe Frank. Data Mining: Practical Machine Learning Tools and Techniques, 2nd ed. San Francisco, CA: Morgan Kaufmann, 2005, 94-96.