Difference between revisions of "Co-training and pairing"

From LearnLab
Jump to: navigation, search
(New page: ---- '''Summary Table''' *Node Title: Learning to read Chinese: Co-training in human (Study 2) *Researchers: Ying Liu, Charles Perfetti, Susan Dunlap, Suemei Wu, Tom Mitchell *PIs: Yin...)
 
 
(One intermediate revision by the same user not shown)
Line 1: Line 1:
 +
Under developing.
 
----
 
----
 
'''Summary Table'''
 
'''Summary Table'''
Line 29: Line 30:
 
== Background ==
 
== Background ==
  
In machine learning research, it has been found that multiple-strategies and multiple modalities facilitate learning (Blum and Mitchell, 1998). However, the effectiveness of the properties of “co-training” theory have not been tested in human learners yet. We carried out this study to directly test two important properties of this theory in human learners. There are two results from the finished experiment and one non-result of interest. Most dramatic is the advantage of written over spoken input. This has nothing to do with co-training but is interesting and important for L2 word learning (translation). Second is the pairs effect, the advantage of spoken + written input presented during unlabelled training compared with either one separately. The independence of the surface features of these inputs (specific speaker, specific font) was not a factor.
+
In machine learning research, it has been found that multiple-strategies and multiple modalities facilitate learning (Blum and Mitchell, 1998). However, the effectiveness of the properties of “co-training” theory have not been tested in human learners yet. We carried out this study as a follow-up study to understand the pairs effect. We have to know whether it is restricted to or larger for [[unlabeled examples|unlabeled trials]].  
 
 
To understand the pairs effect, we have to know whether it is restricted to or larger for [[unlabeled examples|unlabeled trials]]. Experiment 1 did not manipulate pairing in labeled trials. In the fall of 2006, we tested the pairing property under both labeled and unlabeled trails.
 
 
 
To understand the correlation feature better, we are testing the correlation feature in an in-vivo setup with more learning sessions.
 
  
 
== Dependent variables ==
 
== Dependent variables ==
Line 48: Line 45:
 
== Hypothesis ==
 
== Hypothesis ==
  
Pairing of visual font and auditory sound of Chinese characters should enhance learning under both labeled and unlabeled trials, but the benefit is most significant when the trials are unlabeled.
+
Pairing of visual font and auditory sound of Chinese characters should enhance learning more than the unpaired condition, but the benefit is more obvious when the trials are unlabeled.
*
+
 
[[Image:cotraining1.jpg]]
 
  
 
== Findings ==
 
== Findings ==
  
*“Unlabelled paired” trials may aid learning. Learning meanings was facilitated by the addition of unlabeled paired trials that did not provide meaning.
+
[[Image:cotraining3.jpg]]
**However, this unlabeled-trials effect was restricted to cross-modal pairs (spoken syllable and written character); it was absent when only one (spoken syllable) or the other (written character) modality was presented.
 
**Implication: Cross-modal inputs in this situation can establish multiple representations (speech-writing pairs) from which meaning links are more readily retrieved.
 
*Written form learned better than spoken form Large advantage for the presentation of written characters compared with their corresponding spoken syllables for learning a form-meaning pair.
 
*Benefits of uncorrelated examples was not observed.
 
**Correlated examples: Given font and given speaker always co-occur (conditional dependent)
 
**Uncorrelated examples: Given font occurs with all speakers; and given speaker occurs with all fonts (conditional independent)
 
**This is still being assessed by using  multiple learning sessions.  
 
  
[[Image:cotraining2.jpg]]
+
As shown in above figure, the two paired unlabeled conditions had higher accuracies than the two unpaired unlabeled conditions. However, it did not reach statistical significance: the labeled pairing effect was not significant (F(1,6)=0.176, p=0.689), the unlabeled pairing effect was not significant (F(1,6)=2.077, p=0.2), and their interaction was not significant either (F(1,6)=1, p=0.356).
  
 
== Explanation ==
 
== Explanation ==
  
Learning meanings was facilitated by the addition of unlabeled paired trials that did not provide meaning implicates that predictions of the label are generated for unlabeled trials, so they serve as self-generated labeled trials and work as meaningful materials for learning. This effect is especially significant in multiple input situation (paired trials) because the establishment of multiple representations (speech-writing pairs) makes the “label prediction” more accurate.
+
Even though the pairing effect was not statistical significant, the effect showed the pattern as the hypothesis predicted. Pairing of visual and auditory modality is more helpful than any single modality because the visual-auditory connection can be build during the learning process. However, when a label (English translation) is provided, modality pairing does not lead to any benefit because there has already been visual-lexical and auditory-lexical connections for a human learner to process. It might be better for a human learner to focus on one connection at a time due to the working memory load.
 
+
 
== Descendents ==
 
== Descendents ==
  

Latest revision as of 01:01, 12 March 2008

Under developing.


Summary Table

  • Node Title: Learning to read Chinese: Co-training in human (Study 2)
  • Researchers: Ying Liu, Charles Perfetti, Susan Dunlap, Suemei Wu, Tom Mitchell
  • PIs: Ying Liu, Charles Perfetti, Tom Mitchell
  • Others who have contributed 160 hours or more:
  • Graduate Students: Derek Chan, Susan Dunlap
  • Study Start Date Sep 1, 2006
  • Study End Date Dec 31, 2006
  • LearnLab Site and Courses , CMU Chinese Online
  • Number of Students: 20
  • Total Participant Hours for the study: 20
  • Data in the Data Shop: Yes

Abstract

The present study continued to explore how native English speakers learn to speak and read Chinese in a cotraining environment. The experiment consisted of two parts. The first part was training, which was used to teach the input (Chinese fonts and sounds) to output (English translations) mapping of 16 Chinese characters. four training methods were applied in a two by two crossed design. The two factors are labeled pairing and unlabeled pairing. Every subject received all four methods in a counter balanced order. The second part was posttest, in which students produced the English translation when they saw the Chinese fonts or hear the Chinese sounds one by one. The accuracy of translation was recorded.

Glossary

2. A glossary that defines terms used elsewhere in this node but not defined in the nodes that are parents, grandparents, etc. of this node;

labeling; source pairing; source correlation.

Research question

How native English speakers learn to speak and read Chinese under various coordinative learning conditions.

Background

In machine learning research, it has been found that multiple-strategies and multiple modalities facilitate learning (Blum and Mitchell, 1998). However, the effectiveness of the properties of “co-training” theory have not been tested in human learners yet. We carried out this study as a follow-up study to understand the pairs effect. We have to know whether it is restricted to or larger for unlabeled trials.

Dependent variables

Normal post-test: Accuracy of producing the English word under reading and/or listening situation.

Independent variables

Labeling Pairing Variation Correlation

Hypothesis

Pairing of visual font and auditory sound of Chinese characters should enhance learning more than the unpaired condition, but the benefit is more obvious when the trials are unlabeled.


Findings

Cotraining3.jpg

As shown in above figure, the two paired unlabeled conditions had higher accuracies than the two unpaired unlabeled conditions. However, it did not reach statistical significance: the labeled pairing effect was not significant (F(1,6)=0.176, p=0.689), the unlabeled pairing effect was not significant (F(1,6)=2.077, p=0.2), and their interaction was not significant either (F(1,6)=1, p=0.356).

Explanation

Even though the pairing effect was not statistical significant, the effect showed the pattern as the hypothesis predicted. Pairing of visual and auditory modality is more helpful than any single modality because the visual-auditory connection can be build during the learning process. However, when a label (English translation) is provided, modality pairing does not lead to any benefit because there has already been visual-lexical and auditory-lexical connections for a human learner to process. It might be better for a human learner to focus on one connection at a time due to the working memory load.

Descendents

None.

Further information