Harvesting Natural Language and the Language of Knowledge/Power--Presentation From the Folks at Law, Society, and AI: “Information Extraction: Where are we?”, Fabian Suchanek (Télécom Paris)

Pix Credit here

The quantification of law requires a new language.  That new language acquires its operational modalities in code and coding.  But something more is required-- and that 'more' focuses on the liberation of meaning from the cage of text. Text is as much an object of data harvesting as any other object.  Text both serves as a symbolic encasing of meaning, as it serves as the ideological structures that stabilized and rationalizes the meaning (the artifact that is to be extracted) conveyed within the forms of text.  

Pix Credit here

This dual trajectory of datification and of the de-composition of text (and aggregated text within the euphemism-'language') presents challenges and opportunities that have significant effect not merely on communication, but also on (1) the management of communication. and  (2) the instrumentalization of text as symbol-vessels containing harvestable data. In the process, text loses some of its primacy as a means of communication; it gives way to the signification of text which itself now can be communicated by a different set of objects containing meaning. What follows is the fundamental transformation of communication in which the language of managing and curating individuals and social collectives will detach from the language used within social collectives words, images and their textual or pictorial representations). 

Where the operational language of law is grounded in the measurable, and the operationalization of politics is centered on the administration of systems of compliance and assessment, it is only natural that the language of humans (by whatever frm of symbolic representation is is conveyed) becomes data points for the greater project of data driven governance.  To those ends meaning must be extracted from the cage of human to human communication, translated into the language of the measurable by the collective of computing, and then translated against into assessments and judgments that can be (re)conveyed to managed populations.  

Michel Foucault's insights into the disciplinary functions of grimmer in language might now be usually transposed onto the emerging meta-language of knowledge and control: "Expressing their thoughts in words of which they are not the masters, enclosing them in verbal forms whose historical dimensions they are unaware of, men believe that their speech is their servant and do not realize that they are submitting themselves  to its demands. The grammatical arrangements of a language are the a priori of what can be expressed in it." (Michel Foucault, The Order of Things : An Archeology of the Human Sciences (Vintage editions  1994 (1970), p. 297)

That, for me any, formed the background for a marvelous seminar hosted by the folks at Law, Society & AI and its convenors: David Restrepo Amariles (Associate professor, HEC Paris); Michalis Vazirgiannis
Professor (Ecole Polytechnique); Winston Maxwell (Professor, Télécom Paris); and Fabian Suchanek
Professor, (Télécom Paris).

“Information Extraction: Where are we?”, Fabian Suchanek (Télécom Paris)

October 5 @ 14:30 - 16:00
Speaker: Fabian Suchanek (Télécom Paris)
Abstract:

Information Extraction is the task of extracting structured information (such as entities and facts) from natural language documents. It is an important step in making legal documents (such as laws or contracts) available to a computer. In this talk, I will give an overview of how far information extraction has come, what techniques it uses, and what pitfalls it still faces. I will also draw on our own works in the domain of knowledge base construction, rule mining, and the analysis of the power of neural architectures.

Video recording: See the replay
Material: Read the slides

The video and the materials are worth serious study.  Their implication have only now started to become clearer.