Summary
IV. Conclusion §§ 23-24
- §§ 1-6 Introductory remarks
- I. In what sense can computer do so little?
- §§ 7-11 In processing text as text, because our linguistic information is inadequate
- II. Why our philology, up to today, is inadequate to substantiate artificial intelligence in text processing?
- § 12 Incidental factors
- §§ 13-16 Inner reason: for programming a computer our mind needs to analyse micrologically its macro-intuitions.
- § 17 Such reflexive introspection can be done scientifically only by a larger and deeper inductive analysis of natural texts.
- III. The "new" philology
- §§ 18-19 It must be able to formalize the global meaning of a textual set.
- § 20 Some already conquered new philological data.
- §§ 21-22 Trends of research: in sentence types, styles, statistics ...
I feel proud having been invited by Prof. Ott to talk to
your Institute, as your activity is universally considered
as solid, efficient and humble, meaning by this last word that your production is larger than its publicity.
I feel grateful to God, as 30 years ago, precisely on the same day as today, my initiative, already 14 years old, was "confirmed" at the "Kolloquium über Maschinelle Methoden der Literarischen Analyse und der Lexikographie", organized here in Tübingen by Prof. W. Schadewaldt, IBM Deutschland and me. Previously I had reported about it in Bad Nauheim, Oct. 1956 at the annual convention of the Deutsche Gesellschaft für Dokumentation.
§ 1 What I will now say is a cocktail of status artis and testament, as I am 77 years old. Moreover it is a summary of what I have experienced, and not of what has been written by others.
In fact I started to explore how to automatize linguistic analysis in 1946. I started to play with IBM punched cards machines in 1949. I have punched and processed six million cards. I started to use IBM computers as soon as they existed. I have put in I/O more than half a billion records, containing either one line or a word with its "internal" hypertext: of texts of eighteen languages and eight alphabets. I have photocomposed by computer 80,000 pages. I have entered into the computer by optical scanner 12 million characters. Finally, I have compressed, without any loss of information, into 120 million bytes on a CD-ROM the 1,630 million bytes of the Thomistic Latin corpus with its hypertext. I also founded, fifteen years ago, a school of informatics for humanity students at the Catholic University of Milan, the GIRCSE: "gruppo interdisciplinare ricerche computerizzazione segni espressione".
§ 2 Concerning text processing I shall not report either on the hardware or on the software side, but only on the side of the philological analysis of the texts to be processed.
The syntagma "computer and the humanities" refers not only to texts, but also to speech, fine arts, music, mime, theatre, film making, etc.: but for the sake of brevity, I shall talk only of texts, tacitly and analogously implying all the other realms. I have calculated the risk of being qualified as simplistic. I am sure I am not.
§ 3 I am using the word "philology" meaning by it all sciences defining how we speak and how we write.
I use the syntagma "linguistic analysis" attaching to "analysis" its generical value and none of the recent historical specific values. I call "analysis" any census of a text, aimed at detecting and classifying the elements and structures and categories existing in a text.
Deliberately I try to avoid all terminologies specifically adopted by philosophers of language: I do not want to portray a language, but just to draw its lexical map.
§ 4 Thinking about how we do think, speaking about how we do speak, writing about how we do write, is an introspective, reflexive, inner, interior activity: the old classic via interioritatis.
§ 5 In the 1950s newspapers were contrasting the rude and crude technology to the gentle and frail humanity: as if the machine could endanger human thinking. Today specialization leads to a subtler and deeper problem, that of the incommunicability between disciplines, a kind of entropy and decadence of the culture. It invades ecclesiastical sciences too. Splitting the knowledge of man into isolated parts implies breaking the unity of man and of men.
Sciences, humanities, technologies, business, politics, and religious meditation must be composed, not opposed. There is not any human field which is neither spoken of nor written about. All technology is human expression no less than poetry, just aimed differently. Separating human activities, e.g. religion from science, business from ethics, is anti-scientific: it hides some irrational steps, like trying to break one operating system (e.g. a running engine) into two or more machines. Being a priest, people often consider my presence in computer science as exotic, as if you met a camel in your Marktplatz. But it is precisely as a priest that I am doing what I do. In fact analysing texts leads to realizing the presence of the mystery of God at the roots of human understanding and speaking. Moses is on my side: Deuteronomy 30: 11-14 "The word [inviting to chose life and good or death and evil] is very near to you; it is in your mouth and your heart ...". When, July 18, 1956, Norbert Wiener visited me at Gallarate, we agreed about it. Later he published a booklet "God and Golem - Cybernetics impinges on religion" (MIT 1964).
§ 6 I have divided this speech into three chapters:
I. In processing human discourse, the computer can still do very little.
II. Why?
III. Because a new philology is needed.
Frankly, I feel sure of what I am saying, but, as every year I am more aware of how difficult it is to evaluate and administer our personal certitudes, I kindly ask you to listen to me critically.
§ 8 But in processing a text as text, computer science is still detained and entangled in painstaking laboratorial research. I call "text" any "discourse" e.g. an essay, an article, a novel, a patent and even a computer program ... I do not consider as "text" the listings of items, a telephone directory, a train time-table ...
A text is a "system": i.e. an assembling of different components which are adapted and connected to each other so as to constitute a unit which is aimed at operating a definite performance.
In all text there is a global unitarian meaning, i.e. one global, definitely defined and definable architecture of thoughts, resulting from the assembling of as many subunits as there are parts, chapters, paragraphs, sentences, and clauses.
§ 9 As a proof that computer science has not yet subdued texts as texts, I take the fact that there are not any practical large-scale applications of fully automatic indexing and of fully automatic abstracting. "Language industry" will explode the day in which man is able to write programs which are ready to perform. At present, the principle is still being investigated in laborious and laboratorial attempts. Even the history of the "machine translation" confirms this, since the ALPAC Report (Washington DC, 1966) stated that the previous attempts and hasty attacks on it were naïve. It is not the computer which is lacking - memory capacity and programming skills have already been developed for many years - but philology. The linguistic data we are feeding into the computer for text analysis are not yet adequate to the computer potentials and ways of operating. I remember two writings of mine: "The Impact of Cybernetics on the Humanities" (Proceedings of the Jurema 1972 - International Symposion) and "Why Can a Computer Do So Little?" (ALLC Bulletin Vol. 4, 1976, pp. 1-3). The latter was the results of my meetings with the Bundes-Archiv in San Francisco XIV. International Congress of Historical Sciences, August 1975 and later in Koblenz. The former I had read in Zagreb, April 1972.
§ 10 In a word: our philological knowledge is not yet ready to formalize the global meaning of text units and subunits. I will explain it in the next chapter.
§ 11 I conclude this chapter in the following way:
§ 13 I shall now hint at the basic inner reasons. Discourse and text are one end of the operational arc of thinking - speaking. The two extremities are heteroclitous and irreversible. They are one of the realities verifying the mysterious opposition of the inseparable couple: active - passive, author - work, generating - generated: see the copyrights, the royalties, the patents ... Knowing is much more than simply memorizing information. It is also unrestrained curiosity, drafting people to safaris in the jungles of the unknown: but I should say that for our thinking there is not the unknown but the not yet known.
The "sign" per se does not exist in knowledge but in its communication. Speaking is a system, where words are interacting. "A" the transmitter, to "B", the receiver, with "C", the word, transfers "D", the concept or message. A, B, C and D are mutually exclusive. But altogether the four are the transparency between A and B.
The graphic and semantic correlation C + D, sign and sense, word and concept, is free, creative, spontaneous, socially conventional and historically evoluting: not commanded by the nature of the signified nor of signs, but invented and adopted by the social thinking power.
Speech and text are physical entities - strings of signs - linear, closed, fixed, terminated. Thinking (somewhere within our body: who knows where?) is something whose being and acting dimensions are just opposite to the former: generative, multi-dimensional, co-present, having a centre "diffused everywhere", assaulting not only the present, but also - and even more so - the no longer and the not yet ...
Signs of it can be seen in all spoken and written sentences. The elementary unit, the molecule, of our expression is sentence, not word: words are like its atoms. Each sentence is a system, an architecture.
In thinking and consequently in language, two functions or levels emerge: one is the power, the vehicle, the conveyor belt, the active, aggressive, dominating, pushing, logical and criticizing power ...; the second is the messages, the contents, and the "ideas" ...
The former are certainties "with which" we think and speak, i.e. which make us think and speak. The latter are "what we think of and what we speak about".
This statement implies that there are two types or levels of certainties: the vital or generating ones, and the cultural or generated ones: i.e. the power-author ones and the product ones.
§ 14 Semantics and hermeneutics are reflexive sciences of how we are able to understand, i.e. to go to get and receive and grasp the "concepts" or "messages" of another person and consequently the style and mode of his "spirit", from his words and/or from his works, as in all aesthetic interpretation and criticism of music, fine arts, poetry etc. I would not include in it the so-called "de-programming", i.e. when from the performance of a system one tries to reconstruct its program: in fact, the author-work correlation runs between the programmer and the program, but not between the program and its operations.
§ 15 Artificial intelligence on human texts is nothing other than the hermeneutical process delegated by man's mind to a machine in the form of a program. To write such a program, using his inner understanding power, one must first reflexively understand it, i.e. understand how he understands. Moreover, he has to analyse and break such process into its elementary factors and steps, so that he can "express" them in electronic bits.
Our era has been characterized, since the advent of electronics, by the fact that from the sciences of the "macro" we went to the sciences of the "micro", having constructed instruments to penetrate into the recesses of the structure of the matter. In addition and consequently, all computer science is nothing other than the mind demanding for itself to anatomize and mathematically formalize its macro-logic intuitions into its elementary micro-logic units and steps. So, the difficulties of artificial intelligence are in the "intelligence" not in the "artificial": i.e. we do not understand enough how we understand, not enough as to be able to spot our inner logical fibres and steps when understanding ...
Micrological analysis is first of all an exquisitely philological affair.
When we are able to achieve such an analytical, reflexive micro understanding of how we understand, there will be no problem in expressing it both in words and in machine form.
§ 16 It seems that all our inventing and expressing activities start from a sudden and illuminating intuition of a whole new set of various entities and operations producing a definite result. From such intuition we try to discover how to determine one by one all its details.
In reading a text of others, we get first the meaning of at least some words here and there, until we suddenly grasp the global meaning of the set, sentence, chapter etc. From such intuition of the global sense of the whole, we then descend to recognize each individual value in and from its context. In both these "virtuous circles" the basic and starting power of mind seems to be that of dominating at once by a glance the unity of the architecture of the set or system. Such intuition seems to be the essential one in the mental process of hermeneutics.
The global unity of the text is, in any case and certainly, expressed, sometimes redundantly, by the text as a whole i.e. by all its words.
But the human mind is also capable of summarizing (indexing, abstracting, telegraphing), i.e. of comparing the unity of the set with its single components.
All such operations will be programmable when we achieve their micrological analysis. Even if we do not achieve it or not fully, many good computer services will materialize from having tried it.
§ 17 Science is social. Scientific introspection must also then be social. The only possible way I know for such a paradoxical "social introspection", to be effective is what I call linguistic analysis of natural texts.
The "social" always needs the "sign". And the use of the computer not only allows exhaustive censuses, but also gives protection against the dangers occurring when someone builds up philosophies based on personal intuitions only.
Between expressing himself by phonemes and graphemes or by "bits", the only conceptual differences are that computer - the "stupid" machine by which the intelligent mind extends itself - demands to work micrologically and commands a rigidly systematic coherence and consistency of everything.
The global meaning is not purely the aggregate or sum of the meanings of the individual words composing the text, but the "form" of the distribution superimposed to them. The builder of a house has to pay the suppliers for all materials, and also the architect, for the idea and design of the house, though it is not an additional "material".
The textual global meaning is already formalized by all the words existing in machine-readable form. The problem is how to detect in the text, or to insert into it, a few words or other signs characterizing the global unit of the set.
§ 19 The method cannot but be inductive. We have to renew all our linguistic definition and classification, censusing all linguistic elements, one by one, on many very large and different natural texts. Personally I would start doing it on the parts of speech.
§ 20 I shall now summarize some new acquisitions emerged from my works over half a century.
I consider that these facts are already seeds of chapters of the new philology. They are documented in my publications.
All these investigations have been interactive. But the natural force of things always obliged us to do by hand, or better by mind, the first spotting of types and of boundaries between them. The same force has constrained us to start analysing the morphology and semantics of single words in isolation, postponing to a necessarily second stage their syntactical characterization.
§ 21 Finally, I shall hint at some lines of research upon which I have stumbled during the above-mentioned enquiries, and which appear to be necessary for text processing. Some of them are or have been attempted here and there by others, but, if I am not mistaken, all are still laboratorial and tentative.
§ 22 The last paragraph already introduces syntactical analyses. This is a gigantic field, where many spaces are still untouched.
As far as I know, we do not yet have a scientifically documented list of such features.
New interactive methods and strategies of linguistic research are expected. They will be the spring, the engine, the soul of such new philology. Young people find in it enormous quantities of work to which to apply their creative ingenuity.
§ 24 The following pseudo-syllogism give me hope: Computer is the son of man. Man is the son of God. "Ergo" God is the grandfather of the computer ...
aus: Protokoll des 50. Kolloquiums über die Anwendung der EDV in den Geisteswissenschaften am 24. November 1990