Title: Chapter 10Language: Speech Processing
1Chapter 10--Language Speech Processing
The course will not cover chapters 7, 9, 11 or 13
(reduced from course outline). Complete Exam Two
results now available at http//web.uvic.ca/psyc/c
oursematerial/psyc313.m01/313/ExamMarks/Table2
2 Grammar - Continued
- Phrase Structure Grammars
- Definition and example A grammar that represents
sentences hierarchically, with each node of the
hierarchy corresponding to a phrase structure. - Specifies a limited number of sentence parts and
how they can combined. - Can allow for recursion A process can be
recursive if it calls on itself to get its job
done. A definition of something is recursive if
the definition contains the thing defined. For
example, one definition of a sentence is two
sentences joined by the word and . The man
who said S has now arrived - Chomsky (1957) feels they cannot give a complete
account of how we interpret language. E.g, they
have difficulty with the ambiguous sentences
Visiting relatives can be a nuisance. They
are cooking apples The lamb is too hot to eat
The shooting of the hunters was terrible - Transformational Grammar Chomskys system that
can handle ambiguity based on two levels of
representation - Surface Structure A level of representation in
text processing. The surface code refers to the
exact wording and syntax of sentences. - Deep Structure In language, the deep structure
is the representation of a sentence constructed
according to a basic set of phrase structure
rules, without any transformations applied to the
resulting representation. - If transformations are applied, the sentence
might be turned into a question or be phrased in
the passive voice, for example
31. Surface Structure and Deep Structure
- sentences exist at two levels
- the deep structure carries its meaning
- the surface structure is the particular phrase
ordering used to convey that meaning - sentences can have nearly identical surface
structures but have different deep structures - John is easy to please vs. John is eager to
please - sentences can have different surface structures,
but identical deep structures - Sara threw the ball vs. The ball was thrown
by Sara - according to the theory of transformational
grammar we use rules of phrase structure to
generate the underlying idea, or deep structure
of the sentence - then, transformational rules are applied to deep
structure to generate a surface structure
conveying intended meaning
4Psychological Reality of Transformational Grammar
- derivational theory of complexity
- the longer the distance from surface structure
to deep structure (i.e., the more transformations
that are required), the more psychologically
complex the sentence should seem, and it should
be more difficult to comprehend - early returns on this theory were promising but
not all of the results supported it - sometimes the more transformations required
reduced comprehension and sometimes it had no
effect - so the psychological reality of transformational
grammar can be doubted - one answer was to argue that he theory was one of
linguistic competence rather than performance it
describes our knowledge of language, not
necessarily how this knowledge is executed on a
moment-to-moment basis - 3. Evaluation
- the idea that a system of rules was at the heart
of language use was extremely important, and
served as the starting point for just about every
other theoretical approach to language - it is a modular approach language is an innate
and specialized ability - given that the learning environment provides no
help (i.e., is an impoverished stimulus) it can
not be learned it must be innate
5B. The Constraint-Based Approach
- language is learnable, based on the incoming data
and associations - challenge comes not from a return to a
behavioristic approach, but from the neural
network approach to cognition - the associations are not between stimulus and
response, but are associations within neural
networks distributed throughout the human brain - the same neural networks that carry out other
cognitive processes are also responsible for
learning and implementing linguistic knowledge - special modular system neednt be proposed to
explain language learning, use - a non-modular approach
- gradual development and fine-tuning of neural
networks during early linguistic experience play
an important role in rapid learning of language - language is full of the probabilistic constraints
that can be discovered by neural networks and
exploited during the process of language learning - makes use of probabilistic constraints for word
boundaries (discussed earlier) and within
sentences - constraints like these can be easily and rapidly
learned by a neural network through repeated
experience with linguistic strings - given these constraints inherent in the speech
stimulus, it seems possible that language is
learned without use of any special
grammar-learning module
6V. Sentence Planning and Production
- not much research done on processes underlying
speech production - it is difficult experimentally control what
someone talks about or says - A. Stages in Language Production
- sequential steps in the production of
language--conceptualizing, planning, articulating
and self-monitoring - 1. Conceptualizing - deciding what one
wants to say - not much research done on the conceptualization,
as there is no objective way to find out how
ideas come together in speech anticipation - existence of mentalese has been hypothesized a
representational system distinct from language,
from which linguistic expression proceeds - but there is little agreement on its form
- Perhaps a first stage of speaking exists, but
its hard to say much about it - 2 and 3. Planning and Articulating
- formulating linguistic plans organizing our
thoughts in terms of language - then execute plan with articulatory organs
- articulation can be studied through errors
- slips of the tongue reveal speech process
7Self-Monitoring
- self-monitoring processes whereby we keep track
of what were saying, and change it on-line if
necessary - unclear whether we edit what we say before we say
it - but clearly we edit what we have already said
- often we engage in self-repair
- we stop ourselves and correct what weve just
said - self-repairs have a consistent structure
- first, we interrupt ourselves when we detect an
error - second, we issue what might be termed an editing
expression, like um, wait a minute, or
sorry - finally we repair what weve just said by
saying such things as er...I mean - Self-monitoring is arguably not just a final
stage in a serial process, but a real-time
simultaneous corrective
8Slips of the Tongue Their Nature
- slips can occur at any linguistic level phoneme,
morpheme (e.g., suffixes or prefixes), or word - but if an utterance contains a slip, it tends to
be at only one linguistic level - elements within an utterance that interact
- tend to come from similar positions within a word
(e.g., both first parts of words) - tend to be similar to one another (e.g., both
consonants) - tend to receive the same sort of stress they
would have, had they not interacted - tend to be based on phonological, rather than
semantic similarity - Slips obey the rules of phonology
- even when sounds are switched, the resulting
errant word will sound like a word in the language
9Slips of the Tongue Types
- Slips of the Tongue slips of the tongue are a
type of action slip that we did not discuss in
Chapter 4 - Eight Types of Slips
- Shift one speech segment disappears from its
appropriate location and appears somewhere else - Exchange two segments change places (both
segments disappear from their appropriate
location) - Anticipation a later segment replaces an earlier
segment but does not disappear from its
appropriate location - Perseveration an earlier segment replaces a
later segment but does not disappear from its
appropriate location - Deletion leaving something out
- Addition inserting something
- Substitution an intruder replaces an intended
segment - Blend two words combine into one, apparently
because they are both being considered for
selection
10Theoretical Accounts of Slips of the Tongue
- serial account of slips invokes 5 stages
- stage 1 conceptualization determination of
stress patterns determination of syntactical
structure - stage 2 content words and free morphemes are
added - stage 3 bound morphemes are added
- stage 4 function words are added
- stage 5 overt articulation
- stages are independent of one another
- consistent with this assumption, slips seem to
occur only at one level of planning - e.g. a weekend for maniacs (intended a
maniac for weekends) - stress pattern remained the same, as if that was
determined separately - content words switched places but not the bound
morpheme, suggesting the bound morphemes was
added separately from the content word - Note how the sound generated for the stranded
morpheme fits the new context (termed
accommodation) (plural goes from /z/ to
unvoiced /s/) - this suggests that sounds are assembled after
(and independently of) assembly of the words in
the sentence
11More Theories of Tongue Slippage
- parallel processing accounts
- the production of speech exist at a number of
distinct levels (as in serial theory), but these
processors are capable of operating
simultaneously - proposes that words in the lexicon are
represented at four different levels, in terms of
their - phonology, morphology, syntactic roles, and
meaning - processing units at each of these levels work in
parallel, - may excite or inhibit processing at the same
level or at other levels - lexical bias effect
- slips resulting in words more likely than those
resulting in nonwords - true words have morphological representations,
but nonwords do not - activation of these representations would excite
the corresponding processors at the phoneme
level, making errors more likely - this error-inducing feedback would only be
present for words, however, making it less likely
that errors would result in nonwords
12Slips of the Tongue Unconscious Motivation?
- Freudian slip
- Freud thought slips of the tongue (paralepsis)
were windows on a persons fears, anxieties, or
wishes - slips of the tongue mean something they reflect
unconscious motivation - Motley and Baars (1979) -tried to test this
assumption - used a laboratory procedure that induced
participants to make slips, more specifically
spoonerisms, or simple switches between the
initial sounds of nearby words - participants read two-word phrases silently
- every so often, a buzzer sounded and they were to
read the presented pair out loud - to make errors a little more likely, these spoken
trials were preceded by a series of phrases that
made a slip more likely (as in the lexical bias
effect) - to engage participants in a given motivational
frame, the context was manipulated
13Slips of the Tongue Unconscious Motivation?
- one group was preoccupied with shock-related
anxiety - hooked up to (bogus) electrodes and told that
they would be receiving an occasional electric
shock during the procedure - another group was pre-occupied with what might be
termed sexual anxieties - tested by an attractive and provocatively
dressed woman - a control condition also used
- tested with an identical procedure, but with
neither environmental manipulation - the word pairs, when rearranged, referred to
either shock-related or woman-related issues - e.g.--worst cottagecursed wattage share
bouldersbare shoulders - results
- those anxious about shocks were more likely to
make shock-related errors - participants (all male) who were anxious about
sex were more likely to make sex-related errors - slips of the tongue can be made more likely by
contextual variables but most result from a
mis-assemblage of linguistic units
14Conversations and Conversational Structure
- conversations have a fairly stable structure
- usually start with a standard greetings which
usually elicit stock answers - turn-taking speakers alternate overlap is rare
- individual styles of turn-taking vary widely
length of a turn is a fairly stable
characteristic within a given individuals
conversational interactions - standard signals indicate a change in turn a
head nod, a glance, a questioning tone - three principles that (implicitly) guide
turn-taking in conversations - current speaker may choose the next speaker by
directing a comment or question to a particular
person - if this doesnt happen, then it is acceptable for
any person to step into the conversational gap - or, the original speaker may simply continue
talking - these principles are ordered in terms of
priority the first is the most important, and
the last is the least important - Just try violating them in an actual conversation
(but debrief later!)
15Pragmatics
- Pragmatics the practical knowledge we need to
use language effectively - the contribution to meaning of the context of
conversation - relies upon unstated, shared (or unshared!)
knowledge/projection of the situation - meaning that relies on social convention rather
than internal representation depends upon
inferences or implicature - coffee would keep me awake --a request for
coffee or a refusal? - Speakers intend to communicate stuff, listeners
infer what is intended - Grice (1975) identified several maxims that
describe shared assumptions about communication
in context - all support one overriding principle
cooperation - Quantity Say as much as you need to, but not
more - Quality Dont say things that you believe to be
false or for which you lack evidence - Relation Be relevant to the topic at hand
- Manner Be clear avoid obscurity and ambiguity
- Violating these can result in social awkwardness
16Gender and Conversation
- Lakoff (1975) See also
Tannen (1993) - women tend to be more polite as indicated by
using indirect requests -do you
know what time it is? - women tend to employ more tag questions and
hedges in conversation - tag questions are placed at the end of a
statement, as in Its hot out today, isnt it? - hedges are qualifiers, like kind of, or could
be - male conversations tend to feature more
interruption than female conversation - most would agree that there are significant
differences in the ways that women and men
converse, not everyone would agree about what the
differences indicate - may reflect relative differences in power across
women men - May reflect differences in ways of interacting,
rather than powerlessness of women - conversation may be as much a function of the
situation as it is of speaker gender
17Gender Styles In Communication
- Communication is not just about the message (or
content) but also about the way the message is
communicated (manner or style). - Everyones style is unique, but some aspects of
style are widely shared. - One perspective states that there are typical
communication style differences between men and
women. - Typical differences are often confused with
individual differences, and mistaken for absolute
or essential differences (fixed natural kinds or
universal categories). - In fact, typical differences between ways men and
women communicate are not shared by all
individual men and women, and new styles can be
learned while old styles are adapted. - Types are like styles they are not imposed.
- They suggest a pattern, but include an element of
choice.
18Understanding Typical Differences
- Typical differences are not stereotypes.
- They have vague boundaries, and tolerate plenty
of exceptions. - Members of the same type resemble each other but
are not mere duplicates. - Their sameness is more like family resemblances
than the sharing of essence or particular
qualities. - Gender style differences are typical differences
in this sense. - You dont have to worry about being out of style
if you dont fit the type. - The point is not about fitting, but about
increasing understanding of each other. - Just as gender types are not absolute, so human
nature is not about Nature, about fixed givens
and uniformities. - In fact, the same words, gestures, actions can
communicate different or opposing meanings - Examples (see table below)
19Typical Gender Differences
- MEN
- Interpret conversations in terms of hierarchical
social order - Negotiations for status
- Diminish the significance of problems or offer
solutions - Report talk
- Public speech
- Concern for independence
- were separate and different
- WOMEN
- Interpret conversations in terms of connection
and closeness - Negotiate for closeness
- Identify with expression of problems and show
empathy. - Rapport Talk
- Private speech
- Concern for intimacy
- were are close and the same
20(No Transcript)
21Thomson, Murachver, and Green (2001)
- speakers will change language style so that it
converges with the style of the person(s) they
are speaking with - investigated this notion with electronic mail
(e-mail) - each participant communicated via e-mail with two
different netpals, one of whom was ostensibly
female, the other male - in reality, each netpal was the researcher,
writing from a script using either female-style
or male-style language - results
- the nature of what was said (and how it was said)
did not differ with the gender of the
participant - it differed with the (supposed) gender of the
netpal - male and female styles of conversing are not
static characteristics of individuals