Word clusters and the idiomaticity of learner English

1
????????????Word clusters and the idiomaticity
of learner English

• ???
• xujiajin_at_bfsu.edu.cn
• Beijing Foreign Studies University

2
Major points

• Defining word clusters
• Hypothesis and methodology
• DIY step by step
• Data display and interpretation

3
An alternative view of L

• Some linguists see language as an edifice
  assembled with a great amount of ready-made
  templates or building blocks(?????).

4
Some Chinese examples

• ????(Have you had your dinner?)
• ??????(Where are you going?)
• ???(Thats it.)
• ??? (That is to say, kind of)
• ???? (Then, so)
• Such lexical phrases r stored uttered in one
  spurt at a time highly frequent in use.

5
Word clusters
Defining clusters

• A word cluster is a group of words which follow
  each other in a text (Scott 2004). In our case,
  we take a frequency-driven approach to clusters.

6
Aliases of cluster
Defining clusters

• They are similar to phrases in most pedagogic
  grammars, but bear different confusing names in
  the literature, like formulaic sequences, lexical
  bundles, clusters, chunks, multi-word
  expressions, recurrent word combinations,
  pre-fabs, ngrams etc.

7
Word list multi-word list
Defining clusters

• How a chunk is calculated?
• Based on frequency
• Lexical chunks are generated like a multi-word
  list.

8
How a cluster is calculated (e.g.3-word)?
Defining clusters

• The idea of respect comes from the concept that
  everyone, including yourself, has self-worth, and
  therefore should be treated with dignity. Say,
  for example, that you're having a discussion with
  your boyfriend or girlfriend and your opinions
  are different. While you may disagree with each
  other, each of you still has a right to your own
  feelings.

9
How a cluster is calculated (e.g.3-word)?
Defining clusters

• The idea of respect comes from the concept that
  everyone, including yourself, has self-worth, and
  therefore should be treated with dignity. Say,
  for example, that you're having a discussion with
  your boyfriend or girlfriend and your opinions
  are different. While you may disagree with each
  other, each of you still has a right to your own
  feelings.

10
How a cluster is calculated (e.g.3-word)?
Defining clusters

• The idea of respect comes from the concept that
  everyone, including yourself, has self-worth, and
  therefore should be treated with dignity. Say,
  for example, that you're having a discussion with
  your boyfriend or girlfriend and your opinions
  are different. While you may disagree with each
  other, each of you still has a right to your own
  feelings.

11
How a cluster is calculated (e.g.3-word)?
Defining clusters

• The idea of respect comes from the concept that
  everyone, including yourself, has self-worth, and
  therefore should be treated with dignity. Say,
  for example, that you're having a discussion with
  your boyfriend or girlfriend and your opinions
  are different. While you may disagree with each
  other, each of you still has a right to your own
  feelings.

12
How a cluster is calculated (e.g.3-word)?
Defining clusters

• The idea of respect comes from the concept that
  everyone, including yourself, has self-worth, and
  therefore should be treated with dignity. Say,
  for example, that you're having a discussion with
  your boyfriend or girlfriend and your opinions
  are different. While you may disagree with each
  other, each of you still has a right to your own
  feelings.

13
How a cluster is calculated (e.g.3-word)?
Defining clusters

• The idea of respect comes from the concept that
  everyone, including yourself, has self-worth, and
  therefore should be treated with dignity. Say,
  for example, that you're having a discussion with
  your boyfriend or girlfriend and your opinions
  are different. While you may disagree with each
  other, each of you still has a right to your own
  feelings.

14
How a cluster is calculated (e.g.3-word)?
Defining clusters

• The idea of respect comes from the concept that
  everyone, including yourself, has self-worth, and
  therefore should be treated with dignity. Say,
  for example, that you're having a discussion with
  your boyfriend or girlfriend and your opinions
  are different. While you may disagree with each
  other, each of you still has a right to your own
  feelings.

15
How a cluster is calculated (e.g.3-word)?
Defining clusters

• The idea of respect comes from the concept that
  everyone, including yourself, has self-worth, and
  therefore should be treated with dignity. Say,
  for example, that youre having a discussion with
  your boyfriend or girlfriend and your opinions
  are different. While you may disagree with each
  other, each of you still has a right to your own
  feelings.

16
Cluster and idiomaticity
Hypothesis method

• When a word cluster occurs highly frequently, it
  may imply that the cluster is formulaic, being
  able to enhance idiomaticity and fluency (Wray
  2002).

17
Cluster and idiomaticity
Hypothesis method

• Studies in pattern grammar (Hunston 1996), the
  lexical approach (Lewis 1993) to language
  teaching, etc. show that native speakers use much
  more chunks in their language production than L2
  learners do. Alternberg (1998) reported that 96
  of native speakers language follow some
  prefabricated patterns.

18
Cluster and idiomaticity
Hypothesis method

• It is, therefore, believed that the more
  proficient an L2 learner is, the more formulaic
  language is used in his language. Such
  formulaicity makes his language more idiomatic
  and fluent.

19
Measuring the idiomaticity of learner language
Hypothesis method

• Ideally, it would be best to extract a set of
  clusters used in native speakers language, and
  to use the set of clusters to measure the
  idiomaticity of L2 learners language.

20
Measuring the idiomaticity of learner language
Hypothesis method

• However, this is often difficult, as there may be
  thousands of such clusters (Pawley Syder 1983),
  and clusters are often content-related (Schmitt
  Carter 2004). It is not easy to find native
  speakers language data with shared topics.

21
Measuring the idiomaticity of learner language
Hypothesis method

• It is wondered that, if NS data cannot be found,
  proficient L2 learners language data may also
  serve the purpose.
• Word clusters from proficient L2 learners
  language the measure of learner language
  idiomaticity.

22
Data and Methodology

• 90 essays written by university students, with
  scores assigned by expert human raters
• 30 best compositions were chosen from the 90
• Most frequent 3- and 4-word clusters were
  extracted from the 30 best compositions

23
Methodology

• The remaining 60 texts were searched for the of
  clusters found in the 30 best essays
• Statistical analysis was then conducted to see
  whether the of clusters contained in the 60
  compositions correlates with essay scores.

24
Flow chart
No. of clusters in each of the 60 texts
90 essays
60 Others
30 Best
Correlation analysis
PatCount search
Clusters
Ngram list
File-based Concordancing
25
DIY steps

• Cluster/Ngram extraction
• 1. Choose the 30 best students files
• 2. Compute 3-4 word clusters
• 3. Save cluster results

26
DIY steps

• Counting clusters
• 4. PatCount is used to search the 60 texts for
  the 3-4 word clusters.

27
DIY steps

• Correlation analysis
• 5. Numbers of counted cluster frequencies are
  correlated with rater-assigned scores.

28
Data display
29
Correlation co-efficient
???
????
???
???
???
???
0
1
0.8
0.6
0.4
0.2
30
Summary

• Content
• Notion of cluster
• Methodology
• The integration of corpus techniques and
  statistical tests.

31
Some reflections

• In-depth analysis of the results or the
  underlying rationale
• Functional grouping of clusters
• Use of clusters across different proficiency
  levels
• Tag-sequence/POS-gram/colligation

32
Thank you!
33
Assignment

• The correlation between Tag sequence/POSgram and
  idiomaticity