Timo Honkela: Subjects on objects in contexts: Using GICA method to quantify epistemological subjectivity
1 like528 views
The document discusses the GICA (Grounded Intersubjective Concept Analysis) method for quantifying epistemological subjectivity, focusing on the interaction between subjects, objects, and contexts. It highlights the importance of context in meaning and explores methods for capturing subjective evaluations through surveys and data mining. Additionally, it presents case studies on well-being concepts and textual analysis of State of the Union addresses to illustrate the application of the GICA method.
Timo Honkela: Subjects on objects in contexts: Using GICA method to quantify epistemological subjectivity
- 1. Subjects on Objects in Contexts: Using GICA method to quantify epistemological subjectivity Timo Honkela1, Juha Raitio1, Krista Lagus1, Ilari T. Nieminen1, Nina Honkela2, Mika Pantzar3 1 Aalto University (former Helsinki University of Technology) Department of Information and Computer Science (former Neural Networks Research Center, Adaptive Informatics Research Center) 2 University of Helsinki 3 National Consumer Research Center Finland
- 2. Subjects on Objects in Contexts: Using GICA method to quantify epistemological subjectivity Timo Honkela Juha Raitio Krista Lagus Ilari T. Nieminen Nina Honkela Mika Pantzar
- 3. Traditional representation of meaning: Generalized (non-contextual, non-subjective) Gaines: “Designing Visual Languages for Description Logics” http://pages.cpsc.ucalgary.ca/~gaines/reports/KBS/VLL/
- 4. Meaning is contextual red wine red skin red shirt Gärdenfors: Conceptual Spaces Hardin Color for Philosophers
- 5. Meaning is contextual WHITE SNOW - WHITE?
- 6. Meaning is contextual ● “Small”, “big” ● “White house” ● “Get” ● “Every” - “Every Swede is tall/blond” ● etc. etc.
- 7. Learning meaning from context ● Self-Organizing Semantic Maps (Ritter & Kohonen 1989) ● Latent Semantic Analysis ● Latent Dirichlet Allocation ● WordICA ● etc. etc. Honkela, Pulkki & Kohonen 1995
- 8. Learning meaning from context: Maps of words in Grimm fairy tales Honkela, Pulkki & Kohonen 1995
- 10. Meaning is subjective ● Good ● Fair ● Useful A proper theory of ● Scientific meaning has to take ● Democratic this into account. ● Sustainable (opposite to the view given by V. Cherkassky ● etc. about an hour ago)
- 11. Modeling variation of meaning in a community of agents Honkela: ICANN 1993 Steels, Kaplan, Vogt, et al.: Language games (Lindh-Knuutila, Lagus & Honkela, SAB'06) Related to e.g. Steels and Vogt on language games
- 12. Intersubjective Concept Spaces concept symbol concept symbol (shared) space space space space context C1 S1 C2 S2 observations signal Sender d Receiver (Honkela, Könönen, (agent 1) (agent 2) Lindh-Knuutila & Paukkeri 2008)
- 13. Intersubjective Concept Spaces (Honkela, Könönen, Lindh-Knuutila & Paukkeri 2008) Ci: Ndimensional ξ: si ∈ Si → C metric concept An individual space mapping function from symbols to concepts Observing f1 and after symbol S: symbol space, selection process, agent 1 The vocabulary of an φi: Si → D communicates a symbol s* agent that consists of An individual to agent 2 as signal d. When agent discrete symbols mapping from agent 2 observes d, it maps it to some i's vocabulary to the s2 ∈ S2 by using the function φ 11. λ : Ci × Cj → R, i ≠ j signal space D and Then it maps the symbol to some A distance between an inverse mapping point in its concept space by using two points in the φ1 i from the signal ξ2. If this point is close to its concept spaces of space to the symbol observation f2 in the sense of λ, different agents space the communication process has succeeded.
- 14. Gary B. Fogel 11th of June, 2012 WCCI 2012
- 15. GICA: Grounded Intersubjective Concept Analysis Description of the method
- 16. Subjectifying: adding subjective views into object-context matrices Outcome: Subject-Object-Context (SOC) Tensors
- 17. More on subjectification ● A central question in GICA is how to obtain the data on subjectivity for expanding an object- context matrix into the tensor that accounts additionally for subjectivity. ● The basic idea is that for each element in the object-context matrix one needs several subjective evaluations. ● Specifically, the GICA data collection measures for each subject si the relevance xijk of an object oj in a context ck
- 18. Potential sources for subjectification ● Conceptual surveys: ● individual assessment of contextual appropriateness ● Text mining: ● statistics of word/phrase-context patterns ● Empirical psychology: ● reaction times, etc. ● Brain research
- 19. Flattening: unfolding 3-way tensor for traditional 2-way analysis
- 20. GICA: Grounded Intersubjective Concept Analysis Examples of use
- 21. Case 1: Wellbeing concepts ● A conceptual survey was conducted among the participants of the EIT ICT Labs activity “Wellbeing Innovation Camp” that took place between 26th and 29th of October 2010 in Vierumäki, Finland. ● The participants were asked to fill in a data matrix that consisted of the objects as rows and the contexts as columns. ● Each individual’s task was to determine how strongly an object is associated with a context, using Likert scale from 1 to 5
- 22. Data collection CONTEXTS: OBJECTS: Relaxation Happiness Fitness Wellbeing SUBJECTS: Event participants
- 23. MDS: Objects x Subjects Fitness
- 24. NeRV: Objects x Subjects Fitness NeRV: J. Venna, J. Peltonen, K. Nybo, H. Aidos, and S. Kaski. Information Retrieval Perspective to Nonlinear Dimensionality Reduction for Data Visualization. Journal of Machine Learning Research, 11:451-490, 2010.
- 25. SOM: Objects x Subjects
- 26. SOM: Distribution of contexts
- 27. SOM: Contexts
- 28. Case 2: State of the Union Addresses ● In this case, text mining is used for populating the Subject-Object-Context tensor ● This took place by calculating the frequencies on how often a subject uses an object word in the context of a context word ● Context window of 30 words
- 30. SOM: Objects x Subjects
- 31. Analysis of the word 'health'
- 32. Related research and future plans
- 33. Our related research on subjectivity: User-specific difficulty assessment Paukkeri, Ollikainen & Honkela, Information Processing & Management, 2012
- 34. Interoperability ● Current situation: Formalization and harmonization of knowledge representations (e.g. using XML) ● Future possibility: Meaning negotiation between systems based on SOC tensors and further developments Context data is important!
- 35. Enhanced communication, democratic and participatory processes
- 36. Collaboration opportunities ● Theoretical work ● Interdisciplinary: brain research, psychology, sociology, organization research, etc. ● Methodological – Formulation in different theoretical frameworks – Analogical development with crisp>fuzzy: “objective”>subjective ● Experimental ● Case studies ● Research visits, tutorials, workshops ● GICA workshop/summer school in 2013 in Finland