How to describe a language?
- "Shut the door!"
- Sound, order, meaning of individual words
- Different levels.
Phonetics:
- Back to 4000 BC. Without writting.
- How we describe and group together sounds?
- What parts of our mouth to articulate?
- It is independent of the language of the speaker.
Phonology:
- Patterns and variations.
- For example: tpat, ISender, bintlement, zvetsin. Which one can be English?
- Sounds that fit together: decided by humans; and they like to be efficient.
- Find smallest sounds of units that changed in a word, change the meaning.
- E.g. /p/ by /b/ pig → big. Those minimals are called phonemes.
- Hypothesis: many languages will have similar sounds.
- 83 % of some languages have /s/ sound.
Graphetics:
- Physical parts of written symbols.
- E.g. in clay: removed with hard tool.
- In paper: you destroy the pencil but not the paper.
- Materials
Graphology:
- We know how those marks are made, but how those representations have meanings?
- E.g. writting what vs. vhat or that.
Aside on Writing Systems:
- Broca's area is associated with language.
- Vocal apparatus is needed: wind pipe, mouth, etc.
- Where written was started:
- Indus Valley (Indus language)
- Sumerians in Mesopotamia (cuneiform)
- Egyptians (hieroglyphics)
What is a writing system?
- "A system of more or less permanent marks used to represent an utterance [written or spoken] so that it can be recovered more or less exacrlt without ht eintervention of the utterer."
- How many are there? More or less 114.
- Some of them are not decyphered yet.
- Some languages have several writting systems: latin alphabet and cyrilic.
Morphology:
- How do we construct words?
- Small meaningful elements: morphemes.
- dis-agree-ment-s: 4 morphemes / un-happi-ness: 3 morphemes.
- Inflectional
- Derivational
Inflectional morphology:
- Marking differnt forms of one word.
- -s for singular/plural, see/saw for present/past, go/goes etc.
- It is not necessarily additive.
- Does not change the syntactic category of the word (stills a noun, adjective, etc).
Derivational morphology:
- Creating new words with morphemes.
- insitute → institution → institutional → instituionalise...
Syntax:
- No consensus of what constitues a sentence.
- The complete expression of a single thought / consists of a subject and a predicate / ... many definitions.
- Traditionally: sentence → clauses → phrases → words → morphemes.
- Represented in a phrase structure tree. Bottom is concrete, top is abstract.
- Grammar: atalogue of structural descriptions that describe the sentence structure.
- Structural descriptions are grammar rules.
- Prescriptive grammar is learnt in highschool. Helps to normalize the language
- Descriptive grammar allows to understand the rules we follow.
Semantics:
- Study of meaning in language.
- Words name things and their meaning is the thing they name (?).
- In NLP, we do not care about the meaning, but to derive a representation.
Discourse analysis:
- How discourse is created.
- Textual structure:
- Coreference. With pronouns: how do we know what refers to? Analyzing context. What does he refers to?
- Ellipsis. What has been omitted? Where did you see her? Ø Next to the library
- Coherence. Goes back to rhetorical structure theory. Texts should be also coherence.
- Conversational structure. Turn taking (e.g. roblems with Internet communication).
Pragmatics:
- How humans use language in social settings.
- E.g. using social knowledge (What time is it? Well, you have taken a 5 hours nap).