Our Work

Hierarchical Voting Experts: An Unsupervised Algorithm for Hierarchical Sequence Segmentation

Matt Miller and Alex Stoytchev, Developmental Robotics Lab, Iowa State University

Our Work•Extend Voting Experts to more general domains•Use VE for unsupervised segmentation of hierarchically structured sequences•Improve accuracy of segmentation by using“top down” information

Miller and Stoytchev, ICDL 2008

Baby’s task: tupirogolabubedakupadoditupirotupirogolabutupiro… tupiro*golabu*bedaku*padoti*tupiro*tupiro*golabu*tupiro…

[Saffran, Aslin et. al. 1996]

[Cohen and Adams, 2001]Voting Experts’ task: itwasabrightcolddayinaprilandtheclockswere…

itwas*a*bright*cold*day*in*april*andthe*clockswere…

i t w a s a b r i g h t …

1 134 17 6 1Votes:

i t w a s * a * b r i g h t …

Interesting because:1. Entropy metrics very similar to “Statistical cues”

of Saffran, Aslin et. al.2. General model – useful for more than text3. Surprisingly effective, given its simplicity

Low Internal EntropyHigh Boundary Entropy

Experiments•Demonstrate that HVE works•Explore the domain of applicability

“Future” Work•Tokenize an audio stream and apply HVE to find breaks•Use artificially generated and spoken audio•Results on “baby talk” and audio CD data are promising

Miller and Stoytchev, ICDL 2008

Hierarchical Voting Experts: An Unsupervised Algorithm for Hierarchical Sequence Segmentation

Other codes:•Morse code•ASCII Octal•Random Code

3rd Voting Expert:•Improves Accuracy•Top-Down Information

Other Experiments:•No Time