7-8 December 2015
Program: http://www.helsinki.fi/collegium/events/conceptual_change/index.html
Abstracts: http://www.helsinki.fi/collegium/events/conceptual_change/abstracts.pdf
Recommended reading list: https://docs.google.com/document/d/1pFebmjY2Oru2hRKrTl6sE6TXV7CxUfr5rjAb9j1ZIDE/edit?pli=1
- see http://fordhampress.com/index.php/the-architecture-of-concepts-paperback.html
- http://www.jstor.org/stable/10.1086/680201
- System background: http://artsandsciences.sc.edu/engl/clifford-siskin-and-peter-de-bolla-re-enlightenment-future-knowledge
- Concept lab project: https://ccdkconceptlab.wordpress.com/
- http://www.reenlightenment.org/reenlightenment-project
- https://www.youtube.com/watch?v=SoclGz3rCgE
- Methodologies unclear, not documented, word co-occurence tests in different window sizes in order to find out conceptual differences between different texts (e.g. Ny York Times vs. Adam Smith). Based on the examples from the data, small samples and from the results it is difficult to say. Results relying to the shapes.
- Vector semantics (Weaver: Translation (1949/1955), Harris: Distributional structure (1954), Firth: Papers in linguistics (1957))
- Example: early English books corpus
- Word association matrix, used method unclear, the author seems to interpret / explain the results so that things have to have some association (even there might be some issues in the data).
- Subsctraction of the associatied word weights to come up with new words. (horse, feet, inch)
- Not big enough datasets? (Big corpus and diff against the small sample might work better)
- Using document window (what is the relation to paradigmatic and syntagmatic relation?
(break)
Neil Foxlee: From analogue to digital: conventional and computational approaches to studying conceptual change
-
Geschichtliche Grundbegriffe (GG):Basic concepts in history
-
Skinner (other school / discipline)
-
Conceptual change
- (1) change in normative vocabulary (words coming into & going out of use)
- (2A) change in intensity of normative vocabulary
- (2B) Change in direction of normative vocabulary
- (3) rhetorical redescription (extension / widening), novel use of evaluative term
-
MD-CADS: Modern, Diachronic Corpus-assisted Discourse Studies
- Corpus approach: comparing word frequencies between subcorpuses (=good!)
-
GraphColl sofware ( http://www.extremetomato.com/projects/graphcoll ), collocation visualization
- collocation analysis, different metrics
- Analysis of swearing: visualization of 5th order of collocations connected: collocations of collocations..., combining most connected ones as one graph
- SOM clustering
- Word freqency time series analysis
- Language and bodily movement association: http://www.slideshare.net/timohonkela/timo-honkela-multimodally-grounded-ttranslation-by-humans-and-machines
Kimmo Kettunen: Challenges in OCR quality of digitized newspapers – is there a way from data quantity to quality (re)search?
- http://digi.kansalliskirjasto.fi
- no article level in archive, still full pages and text
- downloadable concordance dump 1771-1874 (15G): https://korp.csc.fi/download/Digilib-pub/
- OCR quality improvement
- Human correction (crowdsourcing)
- RE-OCRing
- Post-correction with software
- Usage of mixed methods
Asko Nivala: From the englightenment to romantiscism: Topic modeling the changes of aesthetic discourse in Germany
- A statistical model for identifying the hypothetical "topics" that occur in a corpus
- A topic: a pattern in which the same words occur frequently together (a cluster of words?)
- Mallet, LDA for topic modelling
- Enlightenment corpus (valistus) from 1738-1780 from Gutenberg (in German)
- topic: History, a cluster of words related
- clusters are named by Asko himself, afterwards
- Enlightenment corpus is then compared to romantic corpus topics
- (Methodologically speaking, in the interpretation of the clusters one have to be careful)
- In sampling, window size? In the setting were there only the topic groups presented (and nothing more)?
- (Olli) Applications 1: romanticism / enlightenment classifier, e.g. checking how much a book (or an author) is a representative of a typical romanticism era book?
- (Olli) Applications 2: romanticism generator, romanticism shifter (from a model to another)
- (Olli) Applications 3: diff against two classes
- Comparison against Google Ngram data
- an idea: what about finding the most correlative terms over the time (clustering in time series), instead of taking individual word samples
- Interesting presentation!
- (check Nephological Semantics project)
- formal (trousers) vs. conceptual (definition, meaning) vs. denotational layer (the things, objects itself)
- semasiological perspective: looking words and identifying meanings (top-down, from formal to denotational)
- onomasiological perspective: looking things and how they are conceptualized (bottom-up, from things to words)
- formal onomasiology: synonymy
- conceptual onomasiology: construal (=how thing is conceptualized [to words])
- onomasiological profile: calculate distances between lects, differences between languages or profiles
a setting for finding profiles for words (construals):
- find collocations (co-occurences) [of the words in the study (=migrant, alien)], cluster them by vector space model to find the similarity of collocates. As a result, 12 dimensions of construal (=each cluster forms a dimension, right?). From slides: "identify the strongest collocations of both words; use a vector space method to determine the similarity between those collocates and cluster them"
- then, project the dimensions by the newspaper and year, compare the case study concepts (migrant, alien) over the time and see how their aspect or meaning varies over the time within a dimension, e.g. illegality
- QLVL: http://wwwling.arts.kuleuven.be/qlvl/
- Interesting!
- Mining concepts / words from text, concepts emerging from the text
- Project home: http://www.linguisticdna.org/
(break, discussion 30 min)
- representation of concept change
Katariina Parhi and Jouni-Matti Kuukkanen: Confusing Concepts: Psychopathy and philosophical commitments
- definition on the concept of psychopathy
- practical cases of conceptual change in psychopathy?