language as cognitive process: syntax
terry winograd
addison-wesley, 1983
mostly a book about NLP, CS & linguistics. a prequel to understanding computers and cognition.
1. Viewing Language as a Knowledge-Based Process
rules to understand language (p.8):
- word order rules (syntax)
- vocabulary and word structure
- semantic features (a classification system of which a language user must be aware of)
- reference (e.g. what does it imply? difference between a city and the city)
- time (past, present, future)
- discourse structure (overall coherence)
- attitude messages (e.g. rhetorical question)
- prosody (intonation, inflexion, etc.)
- style conventions
- world knowledge <- this might be interesting, since programming is about world representation
survey of linguistic history:
- prescriptive grammar, or linguistics as law. language as an enforced set of rules, focused on correctness or purity. is mostly wrong (for obv reasons)
- comparative grammar, or linguistics as biology/natural history. language as compared to other languages, to other linguistic structures, grouped in taxonomies, etc.)
- structural linguistics, or liinguistics as chemistry. language as made up of individual, structuring components (saussure)
- generative linguistics, or linguistics as mathematics. language as the result of the combinatorial mental capacities of the speaker (chomsky). the central problem is to define the grammar of a language (the set of procedures which confirm a statement as valid/true).
the view of the book is that language is a communicative process based on knowledge. that’s because programming languages often seem to based on computational linguistics, or language as formal symbol manipulation (which turned out not to be true. or at least not effective compared to ML approaches)
the basic assumption is that there are mental representations in the user which can be represented as data structures in the computer.
if linguistic structure is stratified, eache strate still relates and connects to the stratas around it. this is related to the idea of abstraction
sounds <- -> phonemes <- -> morphemes <- -> words <- -> syntax <- -> representations
(note: in idioms, it seems that words related directly to meaning rather than also word structure)
there is also a difference between:
- long-term knowledge and temporary knowledge
- procedural representation (algorithm based) and declarative representation (data-structure based)
approaches of linguistics can focus on:
which of the above do i choose? (should i even choose? i’m guessing it’s quite obviously text since i’m focusing on source code)
limitations, because language is:
- social
- evocative (aesthetics? def. literature)
- historical (re. social)