Specialization Module

A technical view at communication Shannon-Weaver (1949) model of communication – requirements: model noise source, add redundancy – assumptions: corru...

0 downloads 143 Views 148KB Size
Specialization Module

Speech Technology Timo Baumann [email protected]

Universität Hamburg, Department of Informatics Natural Language Systems Group

Speech as a Communication System

A technical view at communication INFORMATION SOURCE TRANSMITTER

RECEIVER DESTINATION SIGNAL

RECEIVED SIGNAL

MESSAGE

MESSAGE

NOISE SOURCE



Shannon-Weaver (1949) model of communication – –

requirements: model noise source, add redundancy assumptions: corruption of messages is detected, transmitter and receiver are matched

Stratification of the Language System ●



communication is rarely direct e.g. telephoning:

INFORMATION SOURCE TRANSMITTER

RECEIVER DESTINATION

MESSAGE





the higher-level signal becomes a lower-level message layering applies the foundational scientific principle of analysis (Descartes 1637; repr. 1824)

MESSAGE

SIGNAL

RECEIVED SIGNAL

INFORMATION SOURCE TRANSMITTER

RECEIVER DESTINATION SIGNAL

RECEIVED SIGNAL

MESSAGE

MESSAGE

NOISE SOURCE

Conventional division of linguistics into subdisciplines



● ●

area

objective

pragmatics

the study of meaning in context

semantics

the study of meaning

syntax

the study of sentence structure

lexicology

the study of words

morphology

the study of forming words

phonology

the study of a language's sound system

phonetics

the study of speech sounds

is this conventional division correct? That is, does it accord to the best possible analysis? what are the interfaces between the areas? is everything covered by these areas?

Systems theory research on linguistics ●

Serébrennikov et al. (1975) give a detailed account of why the aforementioned layers (strata) are reasonable –



Levelt (1989): psycholinguistic proof for layering –





however, often strong interdependence between layers and the layers' elements (e.g., words are represented by phonemes in a complex way, e.g. based on the words' context) (for speaking: conceptualization, formulation, articulation)

we'll often resort to simple interfaces between layers, when we build technical systems what's better: many simple modules, or fewer complex modules? What are the dimensions of „better“?

Chain model of Communication

find message that describes idea

recover idea described by message

determine structure to convey meaning

determine meaning of structure

sequentialize structure to word stream

recover structure of sequence

represent words through sounds

recombine souds to words

Chain model of Communication

find message that describes idea

recover idea described by message

determine structure to convey meaning

determine meaning of structure

sequentialize structure to word stream

recover structure of sequence

represent words through sounds

recombine souds to words

Chain model of Communication

find message that describes idea

recover idea described by message

determine structure to convey meaning

determine meaning of structure

sequentialize structure to word stream

recover structure of sequence

represent words through sounds

recombine souds to words

A worked example ●

● ● ● ● ● ● ●

A: A: A: A: A: „Could you pass the saltshaker, please?“ channel: [kUd ju: pa:s [email protected] [email protected] pli:z] B receives: „Could you pass the saltshaker, please?“ B: B: B: hands the saltshaker to A.

Small group exercise: Develop and sketch out ways of how the worked example can go wrong. Describe in what ways it's going wrong. What parts of the system are failing?

(Mis-)Alignment within layers ● ●

e.g. you say „Please hand me the salt.“ what's the implication if I respond „Here's the saltshaker.“

Thank you. [email protected]

https://nats-www.informatik.uni-hamburg.de/SLP16

Universität Hamburg, Department of Informatics Natural Language Systems Group

Further Reading ●





Noisy-channel model, chain model of communication: –

Shannon and Weaver (1969): The Mathematical Theory of Communication, paperback edition, reprinted from the Bell System Technical Journal, July and October 1948. The University of Illinois Press. InfBib: T SHA 148 (also in StaBi and on the Web).



M. Pétursson & J. Neppert (1996): Elementarbuch der Phonetik. Buske. StaBi: F Ling 062/6.

Introduction to (German) Linguistics: –

Grewendorf, Hamm and Sternefeld (1989): Sprachliches Wissen. Eine Einführung in moderne Theorien der grammatischen Beschreibung, Suhrkamp. InfBib: A GRE 49762.



... or any other introduction to linguistics; or the relevant chapter in: Jurafsky and Martin (2009): Speech and Language Processing. Pearson International. InfBib: A JUR 4204x.

Systems theoretic views on complex systems in general and on language in particular: –

Serébrennikov et al. (1975): Allgemeine Sprachwissenschaf (translated and edited by Zikmund and Feudel), Volume 2. Akademie-Verlag Berlin. StaBi: A 1974/563: 2.



Bertalanffy (1972): „The History and Status of General Systems Theory“. In: The Academy of Management Journal 15(4), pp. 407-426. via Google Scholar.



Larsen-Freeman and Cameron (2008): Complex Systems and Applied Linguistics, Oxford University Press. StaBi: A 2009 / 7836.

Further Reading ●

Speech Recognition in General: –



Token-Pass Algorithm: –



D. Jurafsky & J. Martin (2009): Speech and Language Processing. Pearson International. InfBib: A JUR 4204x Young, Russel, Thornton (1989): “Token Passing: A Simple Conceptual Model for Connected Speech Recognition Systems”, Tech.Rep. CUED/F-INFENG/TR, Cambridge University.

The Sphinx-4 Speech Recognizer: –

Walker et al. (2004): “Sphinx-4: A Flexible Open Source Framework for Speech Recognition”, Tech.Rep. SMLI TR2004-0811, Sun Microsystems.

Notizen ● ● ●

~100 minutes should prosody be in the slide on linguistic subdisciplines? further parts (

Desired Learning Outcomes ●









students understand the technical model of communication and its limitations students understand Descartes' scientific principle of analysis which leads to stratification as a principle for describing communication systems students know the linguistic subdisciplines that accord to stratification and that don't fit well (e.g. prosody) students understand the problem of modularization and can discuss advantages/disadvantages of particular models students are aware of the many ways that communication can break down (or have smaller problems) and are able to categorize problems and discuss such a categorization