Language Model Adaptation Ekapol Chuangsuwanich 1 Shinji Watanabe 2 Takaaki Hori 2 Tomoharu Iwata 2 James Glass 1 報告者郝柏翰 20130305 ICASSP 2012 ID: 302547
Download Presentation The PPT/PDF document "Handing Uncertain Observations in Unsupe..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Handing Uncertain Observations in Unsupervised Topic-MixtureLanguage Model Adaptation
Ekapol Chuangsuwanich1, Shinji Watanabe2,Takaaki Hori2, Tomoharu Iwata2, James Glass1
報告者：郝柏翰
2013/03/05
ICASSP 2012
1
MIT
Computer Science and Artificial Intelligence Laboratory, Cambridge, Massachusetts,
USA
2
NTT
Communication Science Laboratories, NTT Corporation, JapanSlide2
OutlineIntroduction
Topic Tracking Language Model(TTLM)TTLM Using Confusion Network Inputs(TTLMCN)ExperimentsConclusion2Slide3
IntroductionIn a real environment, acoustic and language features often vary depending on the speakers, speaking styles and topic changes
.To accommodate these changes, speech recognition approaches that include the incremental tracking of changing environments have attracted attention.This paper proposes a topic tracking language model that can adaptively track changes in topics based on current text information and previously estimated topic models in an on-line manner.3Slide4
TTLM
Tracking temporal changes in language environments4Slide5
TTLMA long session of speech input is divided into
chunks Each chunk is modeled by different topic distributionsThe current topic distribution depends on the topic distribution of the past H chunks and precision parameters α as follows:5Slide6
TTLMWith the
topic distribution, the unigram probability of a word wm in the chunk can be recovered using the topic and word probabilitiesWhere θ is the unigram probabilities of word wm in topic kThe adapted n-gram can be used for a 2nd pass recognition for better results.6Slide7
TTLMCN
Consider a confusion network with M word slots.Each word slot m can contain different number of arcs Amwith each arc containing a word wma and a corresponding arc posterior dma.Sm is binary selection parameter, where sm = 1 indicates that the arc is selected. 7
chunk
1
chunk
2
chunk
3
slot
1
slot
2
slot
3
A
1
=3
…Slide8
TTLMCN
8For each chunk t, we can write the joint distribution of words, latent topics and arc selections conditioned on the topic probabilities, unigram probabilities, and arc posteriors as follows:Slide9
TTLMCNGraphical representation of TTLMCN
9Slide10
Experiments(MIT-OCW)
MIT-OCW is mainly composed of lectures given at MIT. Each lecture is typically two hours long. We segmented the lectures using Voice Activity Detectors into utterances averaging two seconds each.10Slide11
Compare with TTLM and TTLMCN
We can see that the topic probability of TTLMCNI is more similar to the oracle experiment than TTLM, especially in the low probability regions.KL between TTLM and ORACLE was 3.3, TTLMCN was 1.311Slide12
ConclusionWe described an extension for the TTLM in order to
handle errors in speech recognition. The proposed model used a confusion network as input instead of just one ASR hypothesis which improved performance even in high WER situations.The gain in word error rate was not very large since the LM typically contributed little to the performance of LVCSR.12Slide13
Significance Test (T-Test)
H0：實驗組與對照組的常態分佈一致H1：實驗組與對照組的常態分佈不一致13Slide14
Significance Test (T-Test)
Significance Test (T-Test)14
Example
X
5
7
5
3
5
3
3
9
Y
8
1
4
6
6
4
1
2