Transcript Corpus Linguistics

Corpus Linguistics
Lecture 1
Albert Gatt
Contact details
 My email: [email protected]
 Drop me a line with queries etc, and
to arrange meetings.
Course web page
 Course web page:
http://staff.um.edu.mt/albert.gatt/home/teachin
g/corpusLing.html
 Details of tutorials, lectures etc will
always be on the web page.
 Readings for the lecture
 Downloadable lecture notes (available after
the lecture)
Suggested text
 T. McEnery and A. Wilson. (2001).
Corpus Linguistics. Edinburgh
University Press
 NB: Over the course of these
lectures, other readings will also be
proposed and made available, usually
online.
Lectures and assessment
 Structure of lectures:
 all lectures will take place in the lab
 usually, about half the lecture (1hr) will
be devoted to practical work
 Course assessment: assignment
 Final essay (ca. 1500-2000 words)
 Essay topics will involve research on
corpora!
Questions…
?
What is corpus linguistics?
 A new theory of language?
 No. In principle, any theory of language is compatible
with corpus-based research.
 A separate branch of linguistics (in addition to syntax,
semantics…)?
 No. Most aspects of language can be studied using a
corpus (in principle).
 A methodology to study language in all its aspects?
 Yes! The most important principle is that aspects of
language are studied empirically by analysing
natural data using a corpus.
 A corpus is an electronic, machine-readable
collection of texts that represent “real life”
language use.
Goals of this lecture
 To define the terms:
 corpus linguistics
 corpus
 To give an overview of the history of corpus
linguistics
 To contrast the corpus-based approach to
other methodologies used in the study of
language
An initial example
 Suppose you’re a linguist interested in the
syntax of verb phrases.
 Some verbs are transitive, some intransitive
 I ate the meat pie (transitive)
 I swam (intransitive)
 What about:
 quiver
 quake
Most traditional grammars characterise
these as intransitive
 Are these really intransitive?
One possible methodology…
 The standard method relies on the linguist’s
intuition:
 I never use quiver/quake with a direct object.
 I am a native speaker of this language.
 All native speakers have a common mental grammar
or competence (Chomsky).
 Therefore, my mental grammar is the same as
everyone else’s.
 Therefore, my intuition accurately reflects English
speakers’ competence.
 Therefore, quiver/quake are intransitive.
 NB: The above is a gross simplification! E.g. linguists
often rely on judgements elicited from other native
speakers.
Another possible methodology…
 This one relies on data:
 I may never use quiver/quake with a
direct object, but…
 …other people might
 Therefore, I’ll get my hands on a large
sample of written and/or spoken English
and check.
Quiver/quake: the corpus linguist’s
answer
 A study by Atkins and Levin (1995) found
that quiver and quake do occur in transitive
constructions:
 the insect quivered its wings
 it quaked his bowels (with fear)
 Used a corpus of 50 million words to find
examples of the verbs.
 With sufficient data, you can find examples
that your own intuition won’t give you…
Example II: lexical semantics
 Quasi-synonymous lexical items
exhibit subtle differences in context.
 strong
 powerful
 A fine-grained theory of lexical
semantics would benefit from data
about these contextual cues to
meaning.
Example II continued
 Some differences between strong and
powerful (source: British National Corpus):
 strong
 powerful
wind, feeling, accent, flavour
tool, weapon, punch, engine
 The differences are subtle, but examining
their collocates helps.
Some preliminary definitions
 The second approach is typical of the
corpus-based methodology:
 Corpus: A large, machine-readable
collection of texts.
 Often, in addition to the texts themselves,
a corpus is annotated with relevant
linguistic information.
 Corpus-based methodology: An approach
to Natural Language analysis that relies
on generalisations made from data.
Example (British National Corpus)
 British National Corpus (BNC):
 100 million words of English
 90% written, 10% spoken
 Designed to be representative and
balanced.
 Texts from different genres (literature,
news, academic writing…)
 Annotated: Every single word is
accompanied by part-of-speech
information.
Example (continued)
 A sentence in the BNC:
 Explosives found on Hampstead Heath.







<s>
<w NN2>Explosives
<w VVD>found
<w PRP>on
<w NP0>Hampstead
<w NP0>Heath
<PUN>.
Example (continued)
new sentence
 <s>
plural noun
 <w NN2>Explosives
past tense verb
 <w VVD>found
preposition
 <w PRP>on
proper noun
 <w NP0>Hampstead
proper noun
 <w NP0>Heath
punctuation
 <PUN>.
Explosives found on Hampstead Heath
Important to note
 This is not “raw” text.
 Annotation means we can search for particular
patterns.
 E.g. for the quiver/quake study: “find all
occurrences of quiver which are verbs, followed
by a determiner and a noun”
 The collection is very large
 Only in very large collections are we likely to
find rare occurrences.
 Corpus search is done by computer. You
can’t trawl through 100 million words
manually!
The practical objections…
 But we’re linguists not computer
scientists! Do I have to write
programs?
 No, there are literally dozens of available
tools to search in a corpus.
 Are all corpora good for all purposes?
 No. Some are “general-purpose”, like the
BNC. Others are designed to address
specific issues.
The theoretical objections…
 What guarantee do we have that the texts in our
corpus are “good data”, quality texts, written by
people we can trust?
 How do I know that what I find isn’t just a small,
exceptional case. E.g. quiver in a transitive
construction could be really a one-off!
 Just because there are a few examples of something,
doesn’t mean that all native speakers use a certain
construction!
 Do we throw intuition out of the window?
Part 2
A brief history of corpus linguistics
Language and the cognitive
revolution
 Before the 1950’s, the linguist’s task was:
 to collect data about a language;
 to make generalisations from the data (e.g. “In
Maltese, the verb always agrees in number and
gender with the subject NP”)
 The basic idea: language is “out there”, the sum total
of things people say and write.
 After the 1950’s:
 the so-called “cognitive revolution”
 language treated as a mental phenomenon
 no longer about collecting data, but explaining what
mental capabilities speakers have
The 19th & early 20th Century
 Many early studies relied on corpora.
 Language acquisition research was based on
collections of child data.
 Anthropologists collected samples of unknown
languages.
 Comparative linguists used large samples from
different languages.
 A lot of work done on frequencies:
 frequency of words…
 frequency of grammatical patterns…
 frequency of different spellings…
 All of this was interrupted around 1955.
Chomsky and the cognitive turn
 Chomsky (1957) was primarily responsible for the
new, cognitive view of language.
 He distinguished (1965):
 Descriptive adequacy: describing language, making
generalisations such as “X occurs more often than Y”
 Explanatory adequacy: explaining why some things
are found in a language, but not others, by appealing
to speakers’ competence, their mental grammar
 He made several criticisms of corpus-based
approaches.
Criticisms of corpora (I)
 Competence vs. performance:
 To explain language, we need to focus on
competence of an idealised speaker-hearer.
 Competence = internalised, tacit knowledge of
language
 Performance – the language we speak/write – is
not a good mirror of our knowledge
 it depends on situations
 it can be degraded
 it can be influenced by other cognitive factors
beyond linguistic knowledge
Criticisms of corpora (II)
 Early work using corpora assumed that:
 the number of sentences of a language is finite (so
we can get to know everything about language if the
sample is large enough)
 But actually, it is impossible to count the number of
sentences in a language.
 Syntactic rules make the possibilities literally infinite:
the man in the house (NP -> NP + PP)
the man in the house on the beach (PP -> PREP +
NP)
the man in the house on the beach by the lake
…
 So what use is a corpus? We’re never going to have
an infinite corpus.
Criticisms of corpora (III)
 A corpus is always skewed, i.e. biased in
favour of certain things.
 Certain obvious things are simply never said.
E.g. We probably won’t find a dog is a dog in our
corpus.
 A corpus is always partial: We will only find
things in a corpus if they are frequent
enough.
 A corpus is necessarily only a sample.
 Rare things are likely to be omitted from a
sample.
Criticisms of corpora (IV)
 Why use a corpus if we already know things
by introspection?
 How can a corpus tell us what is
ungrammatical?
 Corpora won’t contain “disallowed” structures,
because these are by definition not part of the
language.
 So a corpus contains exclusively positive
evidence: you only get the “allowed” things
 But if X is not in the corpus, this doesn’t mean
it’s not allowed.
 It might just be rare, and your corpus isn’t big
enough. (Skewness)
Refutations
 Corpora can be better than introspectvie
evidence because:
 They are public; other people can verify and
replicate your results (the essence of scientific
method).
 Some kinds of data are simply not available to
introspection. E.g. people aren’t good at
estimating the frequency of words or structures.
 Skewness can itself be informative: If X occurs
more frequently than Y in a corpus, that in itself
is an interesting fact.
Refutations (II)
 By the way, nobody’s saying “throw
introspection out the window”…
 There is no reason not to combine the corpusbased and the introspection-based method.
 Many other objections can be overcome by
using large enough corpora.
 Pre-1950, most corpus work was done manually,
so it was error prone.
 Machine-readable corpora means we have a
great new tool to analyse language very
efficiently!
Corpora in the late 20th Century
 Corpus linguistics enjoyed a revival
with the advent of the digital personal
computer.
 Kucera and Francis: the Brown Corpus,
one of the first
 Svartvik: the London-Lund Corpus,
which built on Brown
 These were rapidly followed by
others… Today, corpora are firmly
back on the linguistic landscape.
Summary
 Introduced the notion of corpus and
corpus-based research
 Gave a quick overview of the history
of this methodology
 Looked at some possible objections to
corpus-based methods, and some
possible counter-arguments
Next lecture
 We look more closely at some
important properties of a corpus:




Machine-readability
Balance
Representativeness
…