Home > Department of Methodology > Seminar series > 3rd MI Conference on Text Mining Methods (TMM)


3rd MI Conference on Text Mining Methods (TMM)

Theme: Validating Evidence

London School of Economics and Political Science,

Monday & Tuesday 9-10 July 2012

Dr. Aude Bicquelet a.j.bicquelet@lse.ac.uk
|Professor Martin W. Bauer M.Bauer@lse.ac.uk
Any social research method is ‘representational’ in nature and therefore has to face the question of validity. The key of the validity problem in text-mining, as well as with other research methods, is that research should represent ‘truthfully’ its topic of investigation. Established social research methods have a long standing debate on issues of validity of data and of conclusions drawn from data, the booming and relatively new field of text mining is still seeking to establish its ‘language game’ and validation procedures. The raison d’être of text-mining is the absence of direct observational evidence, thus text miners face substantive, empirical and methodological obstacles when trying to apply traditional methods of validation to their research.

The aim of this 3d LSE-MI Conference on Text Mining Methods (TMM) is to bring together social researchers working with textual data and to provide them with a platform to discuss:
(a) The challenges arising from validation issues.
(b) Methods to reduce risks of making unwarranted inferences from text data.

Presentations will focus on corpus construction, coding, indexing, text classification and their quality indicators designed to provide compelling reasons for taking text-mining analyses seriously. Talks will feature applications of established and innovative approaches to textual data with emphasis on discussing their take on validity (face, social, sampling, semantic, structural, functional, correlative and predictive). We will provide an opportunity to gauge the efforts of text miners in the widest sense to ensure that their research withstands the test of independent evidence and are thus able to yield sound conclusions to inform decisions.