A procedural model of language based on encoding

A procedural model of language based on encoding and selection out of a communicative field Dorota Zielińska Kraków, Poland Dorota. Zielinska@yahoo. co. uk

A developmental perspective on language Languages regarded as abstract semiotic systems do not change by themselves - only speakers do, and consequently, speakers change their linguistic performance. (after Bunge, 2003) Explanatory models of language cannot concern abstract structures alone. Given the functioning of the brain, models of language must have a developmental character.

Encoding and selecting in language use and constitution Encoding: When a speaker identifies the referent of a linguistic symbol, an existing correlation between a form and a potential reference (its representation) is strengthened (and possibly modified) by strengthening the relevant synapses (Hebb’s model of memory). Selecting: A speaker uses the encoded content related to the symbol being interpreted to select out of the set of representations of viable items, the one that resembles the encoded item best. The representation of the item selected is added to the current representation of the given symbol through strengthening relevant synapses. The set of viable items is called a communicative field.

Communicative field (c-field) n C-field: a set of ”expectations, associations” (eventually verbalized), each with assigned probability of its occurrence. The fridge is empty. I am going to |……. | Pragmem 1: shopping for food buy/get/bring bread, butter, milk, meat, etc. at (possible shops typical for that person, for the given community, in a given situation). Verbal expressions of that content, Pragmem 2: eating out C-field needs a formal model, perhaps Peter Bruza’s?

Selection – can be modelled by a type of supervised learning technique, e. g. , analogical modelling (Skousen, 1992). It should depend on the number of the tokens of a given type used recently, the total similarity of tokens to the token considered, the probability of recalling a given type in a given context. Selecting can be compared to using 2 points to select exactly one among a set of lines

Selection n The task is to categorize item A as either category Y or category X (find the distance between A and the sets Y and X, respectively) Y Y X A X X X x X The axes represent two selected parameters describing the items X, Y, A

Selected problems of truth conditional models of linguistic meaning n n Basic encodings cannot be transferred between individuals. (due to postulating selection there is no need to share a prototype) No good models of lexical categories. (Aristotelian, a fuzzy set, prototype centred, family resemblance) models. No satisfying way of modelling usage of language (eg. , compositionality, metaphor, ellipsis); linguistic change, language acquisition

Examples of such problems and solutions within the framework advocated n n blue eyes, red eyes 1 , red eyes 2 , Mary has bought a lot of flowers for her hotel last winter. The apocalypse has every right to scare us. While I am growing older and older, my children are growing older and I am having problems lifting them. „Are you 18? ” ”Yes, I am”- replied a 20 year-old man.

Meta-informational structure of a message as the result of selection What language is spoken in Burma? English is spoken in Burma. {[out of the languages spoken in Burma] English} n Where is English spoken? English is spoken in Burma. (B. Hajcova, Partee, Sgall) {English [out of all languages spoken in the world} n n Many people read few books. (different few books each) Few books are read by many people (Palmer) (the same few books, e. g. the Bible) Wielu ludzi / czyta / niewiele książek. Niewiele książek/ czyta/ wielu ludzi.

Selective usage of language and dictionary definitions n The dachshund is 200 years old. The dachshund is 2 years old. dachshund [breed of a dog] [dog of a breed] This is big. This is a whale. (? ) This is a big whale. (Chierchia) n {This [thing]} is big. {This [animal]} is big. {This [whale]} is a big whale n Mary is John’s wife {[a woman] a man is married to}. Mary is John’s wife (vs. a sister, mother, daughter, friend). n This is a boxing competition for boys {young [males]}. This is a school for boys {male [children]}. This is a school for boys {young, male [humans]}. +++++++++++++++++ Chess expert explains chess to children. (taken out of a book catalogue) {chess expert[book’s author]}

Fuzzy (probabilistic) metainformational structure of a sentence? What has been negated? „Pragnę Państwa uspokoić, że nie stwierdzono trotylu na wraku tupolewa” (Attorney General of Poland , a press conference 25. 10. 2012) I would like to calm you down, no TNT has been found/ (attested) on the Tupolev.

Developing a new tool for assessing the meaning conveyed - also for judicial purposes 1. It has not been attested that p. 1. 2. p = there was trotyl the assessment (that p) has not taken place yet, has not been completed, whatever the status of p and whether we are aware of its truth or not; P is not true. The assessment has been completed, but there was no trotyl and that’s why it has not been attested (a structurally motivated interpretation) 2 we still do not know the value of p (the interpretation implied structurally – one declared as intended by the speaker) 3. It has been attested that there was no trotyl. (actual interpretation indicated by the results of the research, supported by the rest of the text. ”We would like to calm you down, …. , +description of the test procedures. ) A dictionary of the conveyed meaning of phrases. Experiment: corpus taken phrases, asking subjects to choose the most plausible interpretation, and independently, about the plausibility of p. Eventually, purely automatic assessment

The benefits n n n Categorization (no need to establish and share prototypes, no need to consider the allowable degree of departure from a prototype) Language efficiency (zmień nogę /change a leg/ – (when cantering vs. - when posting a trot) Metaphorical use of language ellipsis Language acquisition is possible without explaining the meaning of every word, the meaning of new grammatical constructions can be inferred (and it happens gradually) Language change, self-organization and self-regulation is possible (and it happens gradually)

Mechanisms of self-organization and self-regulation in language n Local optimizing processes (reductionist, biopsychological) Local regularization of meaning and form correlation during categorization (language use) (ganging effect in Skousen) n Global optimizing processes (systemic, biopsycho-social) A large chunk of the source of data of a given speaker used in the selection process (the comparison set) is made up of utterances pronounced by successful speakers. Successful speakers use an efficient variety of language, which helped them succeed in life. Their speech constitutes input for a large number of people and thus their standards are replicated.

Biological, psychological, and linguistic motivation n n n Two hemispheres and their predominant operating modes Fast mapping experiments and the speed of acquiring vocabulary in early childhood Forming predictions (eye tracking experiments, false memories, scripts etc. ) Only about 10%-20% of the information conveyed with natural language is coded. (see also Pareto laws, etc. ) ”word entanglement” experiments (Peter Bruza) Priming effects (Firth J. R. (1934) ”you shall know a word by a company it keeps” The likelihood of options

Functional organization of the (discrete) communicative field n n n systemism cum emergergentism mice experiments (nests and the organization of neural cliques: Li, Tsien) basic encodings ”Mey (2001) pragmemes”- like fields at the bottom of the organization?

Bibliography n n n Zielinska, 2007 Proceduralny model języka. WUJ. Zielinska, D. 2007. The selective mode of language use and the quantized communicative field. Journal of Pragmatics. 39, 813 -830 Zielinska, D. (to appear 2012) Utterance and sentence meanings from the perspective of theory of models in empirical sciences. In: Perspectives in Philosophy, Pragmatics and Psychology. ed. Alessandro Capone, Franco Lo n Piparo, Marco Carapezza, Springer. Zielinska, D. Science of Language in the Language of Science. (in preparation, available as a file from the author)

Beneš, Martin, 2012 n n n K problému kompozicionality jazykového významu ; benes at ujc dot cas dot cz; Ústav pro jazyk český AV ČR, Prague, Czech Republic The first part of this article documents the fact that all four possible combinations of the Czech adjectives železniční (rail) and silniční (road) and nouns podjezd (underpass) and nadjezd (overpass) can refer to both of the two possible kinds of grade-separated road/railway crossings. The second part of the article discusses this particular fact in connection with the notion of compositionality and interprets it using the procedural model of language constructed by the Polish linguist Dorota Zielińska. It is argued that compositionality cannot be viewed as a phenomenon defined in a binary manner, but rather, that being (non-)compositional is a matter of degree.

Thank you