From mathematical classification to thematic analysis of philosophical texts

Introduction
Since the 1970’s, research in cognitive sciences and information technologies (IT) has had an important impact on the reading and analysis of humanities texts. Various artificial intelligence classification and categorization strategies, have been used to assist description and interpretation of texts. Rules and concepts are predefined and applied systematically to a text. It is a top-down analysis process. This type of strategy still highly influences what is called “qualitative analysis” (Glaser & Strauss : 1967), “Grounded theory”, “Computer assisted qualitative data analysis” (Barry : 1998, Alexa & Zuell : 1999a) or, in more general terms, “Computer assisted reading and analysis of text (CARAT)”. Nevertheless, “Qualitative research usually means the collection and analysis of unstructured textual material in order to develop concepts, categories, hypotheses, theories. Thus, most of the time during “qualitative data analysis” is spent on reading, rereading, interpreting, comparing and thinking on texts.” (Kelle : 1997a).
On the other hand, more and more bottom-up strategies are appearing for inductive text classification and categorisation. These strategies are traditionally applied to information retrieval problems, but their power can also be used in computer assisted reading and analysis of text (CARAT) for lexical analysis, automatic generation of hypertext links, knowledge extraction and thematic analysis. In this paper, we present an application of these classification and categorization technologies to one of these fields : the thematic analysis of philosophical texts.

Classification
Text classification can be defined as an operation, that, when applied to texts described by some of their characteristics, builds equivalent classes on them : “Text classification is the automated grouping of textual or partially textual entities” (Lewis & Gale : 1994).
It is important to note that classification is usually not applied to entire texts but only to fragments of texts described by their “representors” (specific words, abstracts, summaries, etc.). Secondly, classification is not an end in itself. It is a first step in a complex cognitive process of computer assisted text interpretation (Meunier : 1996). Concretely, this means that text classification must link to the dynamic process of text interpretation and hence must be appreciated according to this end.

Methodology and experimentation
Our thematic analysis process is realized through 5 distinct steps : 1) The text is prepared, 2) it is then transformed into a vectorial model, 3) a classifier is applied, 4) thematic links are extracted and 5) results are interpreted.

Text preparation
To be effective, text classifiers need to define two different types of entities : the Fragments of texts (FRAGS) and the units of textual information (UNIFS). Although quite simple to understand, this deconstruction of a text into FRAGS and UNIFS relies on many implicit postulates : some of them are linguistic in nature, others are mathematical.
In this experiment, the textual data was a philosophical text : Descartes’ Discours de la Méthode. This untouched text (except for basic printing noises) contains 21 436 words. In this case the UNIFS were the lemmatized words left after eliminating hapax and functional terms. For this preparation we used a in-house program called CONTERM.

Text transformation into a vectorial representation
From then on, the text is transformed into a vectorial model (Salton : 1989). This procedure requires that each vector represents a segment and describes its units of information (words, n-grams, etc.). From these vectors a matrix is built (FRAGS, UNIFS). The values given to each entry of the matrix depends on the model chosen (e.g. presence, absence, fuzziness, weighting, etc.). On this textual matrix, we now have to define the core of the classification operation; that is, the discriminant function. It is on this matrix that classifiers are applied.

The classifying application
In the literature of mathematical text classifiers, many types of classifiers have been proposed. All of which have their parameters; hence, fecundity and limits. The most common ones are statistically oriented (clusterizers, correlators, factorial analysis (Reinert : 1994), principal component analysis (Benzecri : 1973)). One successful implementation of these types of models has been, in the information retrieval community, the SMART system of Salton (Salton : 1989) or some variation of them such as the one found in Latent semantic approaches (Deerwester : 1990). Also, some more probabilistic techniques have been tested, such as Bayesian classification, Nearest neighbor (Hart : 1998), Neural Networks classifiers (Kohonen : 1982, Anderson : 1976), genetic algorithm (Holland : 1975), Markovian fields classifiers (Bouchaffra & Meunier : 1995b). All of these mathematical models have been applied to textual information processing.
In this experiment, we have chosen the neural ART classifier(Carpenter & Grossberg : 1991). We here explore this classifier and study its relevance for the CARAT processing, focusing on its application to thematic analysis of philosophical corpus. The purpose of this experiment is not to demonstrate its validity for texts classification, as it is often done in classical IR experiments, but instead to show its fecundity in philosophical studies.
In this ART classifier, the discriminant function measures the difference of the input segments by their weighted features (words) but through a competition amongst the segments. That is to say, the fragments that belong to a set are the winners of their weight competition. In most models, this computation consists in accepting a segment only if the impact of all the features of an input attains a certain threshold (called RHO).
The ART I classifier (Grossberg : 1988) is highly interesting. It allows plastic processing of information. When a new pattern is added, ART system classifies it in the set of classes formed by the means of a competition process undertaken during the training phase. The essential principle of this model is based on the interaction of two levels of neurons that enter into a resonance phase such as shown in the following figure.

The system receives in its first level (N1) the text fragments which are sent, after a modification of their distributed weights, to the second level of neurons (N2). This transmission implies various complex differential operations, variation of the activating forces, degradation, shunting, etc.
This second level will then possess many such modified patterns. Each of which serves as prototype for new-comers at the first level. Hence, each new input will be compared to the prototype according to a resonance criteria RHO. If the correspondence is positive then the input is entered in the class of the prototype. If not, it will be considered as an emergent prototype. The adaptability emerges by this constant modification of the levels’ interconnections. And this new emerging prototype will then serve for starting up a new class. As the learning goes on, there will be a consolidation of this resonance.

At the end of this process, a set of classes of segments is produced. And from each class of segments, the lexicon is extracted. We then have for each class a specific lexical list. It is on these classes that the thematic analysis begins.

Thematic extraction
This analysis starts by a preferred “thematic word” that can be found in one of the segments of the classes produced. In this experiment we have started with the word (or concept) “reason”. From this word, thematic relations are then identified with other thematic words belonging to other classes. In the sample results shown below, a particular analysis of Descartes’ Discours de la Méthode starts with the word “reason” found in the class 1. This class explores the metaphysical question of the human existence as found in Descarte’s philosophy. But, by the same token, we also observe that the same word (reason) operates semantically in other contexts. Accordingly, the reader could then decide to discover (by exploring other classes where the word “reason” is also found) many other different fields of Descartes’ writing : the mind (class 4), the knowledge (class 38), the biology (class 66). According to this view, the reader is able to “direct” his analysis according to his own reading interests. With the help of the preliminary classification function, he can explore many different themes found in Descartes’ philosophy. This hypothesis or strategy is then applied to the entire text.

Results and analysis
The results have shown that many thematic links could be relevant for the reader. Here are some results obtained after the classification process :

Also, these results, after being compared with the traditional Descartes’ commentators (Rodis-Lewis : 1966 and 1985, Gueroult : 1953, Laporte : 1950, etc.), were similar with the classical interpretations of Descartes’ philosophy for this particular theme.

Evaluation of classifiers for thematic analysis
From an epistemological point of view, this classification method usually relies on a benchmark produced by a set of objective experts (judges). Results are compared to these benchmark results. But we believe that reading and analysis of texts remains a subjective process. Therefore, this supposed "objective" evaluation methodology seems difficult to apply because, the discovery of the semantic content of a text is always a complex cognitive process that is guided by the reader’s personal reading interests. For each reader, and mainly in philosophical texts, there is a personal interpretative path. Nevertheless, the use of such classification methods gives an objective orientation to the analysis. It does not impose any kind of strategy to the reader. But offers him a multitude of possible paths.
One possible way to evaluate the quality of a computer thematic classification would be to compare it to various best practices or to offer the user some type of relevance feedback measures produced by HMM strategies or genetic algorithms.

References
Alexa, M. & C. Zuell. (1999a). Commonalities, difference and limitations of text analysis software: The results of a review. ZUMA arbeitsbericht, ZUMA : Mannheim.
Anderson J. R.. (1976). Language, Memory and Tought. New York : John Wiley & Sons.
Barry, C.A. (1998). Choosing qualitative data analysis software : Atlas/ti and Nud*ist compared. Sociological research online 3(3).
Benzecri, J.-P. (collab). (1973). La Taxinomie. Vol. I. l'analyse des correspondances. Dunod, Paris.
Bouchaffra, D. & Meunier, J.-G.. (1995b). A Thematic Knowledge Extraction Modeling through a Markovian Random Field Approach. In the 6th International DEXA 95 Conference and Workshop on Database and Expert Systems Applications, Sept. 19-22, London, UK.
Carpenter, G. & Grossberg, G. (1991). An Adaptive resonnance Algorithm for Rapid Category Learning and Recognition.
Deerwester, S., Dumais. S. T., Furnas, G. Landauer. T. K. Harshman. (1990). Indexing by latent semantic analysis. In Journal of the American Society for Information science, pp. 391-407.
Glaser, B.G. & Strauss, A.L. (1967). The discovery of grounded theory. Strategies for qualitative research. Chicago : Adline.
Grossberg, S. (1988). Neural Network and Natural Intelligence. Cambridge : MIT Press.
Gueroult, M. (1953). Descartes selon l’ordre des raison. Paris : Aubier.
Hart, P. E. (1968). The condensed nearest neighbor rule. IEEE, Trans. on Informations theory. IT. 14 : 515.
Holland, J. (1975). Adaptation in Natural, and Artifial Systems. University of Michigan Press. Ann Arbor, Michigan.
Kelle, U. (1997a). Theory Building in Qualitative Research and Computer Programs for the Management of Textual Data. In Sociological Research Online, 2 (2)
Kohonen, T. (1982). Clustering, taxonomy and topological Maps of Patterns. In IEEE Sixth International Conf. Pattern Recognition, pp.114-122.
Laporte, J.M.F. (1950). Le rationalisme des Descartes. Paris : PUF.
Lewis, D.D. & Gale, W.A.. (1994). A sequential algorithm for training text classifiers. In W. Bruce Croft & C. J. van Rijsbergen (eds.). Proceedings of the seventeenth annual international ACM-SIGIR conference on research and development in information retrieval, pp. 3-12. Dublin : Springer-Verlag.
Meunier, J. G. (1996). La théorie cognitive: son impact sur le traitement de l'information textuelle. In V. Rialle et Fisette, D. (Eds. ). Penser l’Esprit, Des sciences de la cognition à une philosophie cognitive, pp. 289-305. Grenoble : Presses UG.
Reinert, M. (1994). Quelques aspects du choix des unités d'analyse et de leur contrôle dans la méthode Alceste. In L. L. S. Bolasco & A. Salem (eds.). Analisi Statistica dei Dati Testuali, vol. 1, pp.19-27. Rome : CISU.
Rodis-Lewis, G. (1966). Descartes et le rationalisme. Paris : PUF.
Rodis-Lewis, G. (1985). Idées et vérités éternelles chez Descartes et ses successeurs. Paris : J. Vrin.
Salton, G. (1989). Automatic Text Processing. Addison Wesley.