Latent Dirichlet Allocation based organisation of broadcast media archives for deep neural network adaptation

Doulaty, Mortaza; Saz, Óscar; Ng, Raymond W. M.; Hain, Thomas

doi:10.1109/asru.2015.7404785

Cited by 10 publications

(17 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…All of the in-domain data was used for training the aLDA model with the procedure described in section 2 using a vocabulary size of 1024 (number of Gaussian mixture components) and 2048 latent domains. Both these values were selected based on our previous experiments [4,5]. The trained aLDA model was then used to get the posterior Dirichlet parameter γ for all of the utterances in the training, dev and test set.…”

Section: Alda Data Selectionmentioning

confidence: 99%

“…* Core part of this work was performed while the author was studying at the University of Sheffield In this paper we propose to use acoustic Latent Dirichlet Allocation (aLDA) for matching acoustically similar data to the limited in-domain data from a pool of diverse data. aLDA is already applied for domain discovery [3] and domain adaptation [4] in automatic speech recognition as well as media entity recognition, such as show and genre identification in information retrieval systems for media archives [5,6,7].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Latent Dirichlet Allocation Based Acoustic Data Selection for Automatic Speech Recognition

Doulaty

Hain

2019

Interspeech 2019

Self Cite

View full text Add to dashboard Cite

Selecting in-domain data from a large pool of diverse and outof-domain data is a non-trivial problem. In most cases simply using all of the available data will lead to sub-optimal and in some cases even worse performance compared to carefully selecting a matching set. This is true even for data-inefficient neural models. Acoustic Latent Dirichlet Allocation (aLDA) is shown to be useful in a variety of speech technology related tasks, including domain adaptation of acoustic models for automatic speech recognition and entity labeling for information retrieval. In this paper we propose to use aLDA as a data similarity criterion in a data selection framework. Given a large pool of out-of-domain and potentially mismatched data, the task is to select the best-matching training data to a set of representative utterances sampled from a target domain. Our target data consists of around 32 hours of meeting data (both farfield and close-talk) and the pool contains 2k hours of meeting, talks, voice search, dictation, command-and-control, audio books, lectures, generic media and telephony speech data. The proposed technique for training data selection, significantly outperforms random selection, posterior-based selection as well as using all of the available data.

show abstract

Section: Alda Data Selectionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Latent Dirichlet Allocation Based Acoustic Data Selection for Automatic Speech Recognition

Doulaty

Hain

2019

Interspeech 2019

Self Cite

View full text Add to dashboard Cite

show abstract

“…Kim et al [8] used the whole shows to train the LDA models and used the domain posteriors as features for an SVM classifier. In this work we followed our previous setup [18,21] where only speech segments are used to train the LDA model. For each show, the domain posteriors of its segments were accumulated and length normalised and used as features for the discriminative classifier in the later stage: It is important to note that this dataset is by orders of magnitude larger than most of the datasets used in the literature for the genre ID task [2,4,5,6,8,15].…”

Section: Acoustic Latent Dirichlet Allocationmentioning

confidence: 99%

Automatic Genre and Show Identification of Broadcast Media

Doulaty

Saz

et al. 2016

Interspeech 2016

Self Cite

View full text Add to dashboard Cite

Huge amounts of digital videos are being produced and broadcast every day, leading to giant media archives. Effective techniques are needed to make such data accessible further. Automatic meta-data labelling of broadcast media is an essential task for multimedia indexing, where it is standard to use multi-modal input for such purposes. This paper describes a novel method for automatic detection of media genre and show identities using acoustic features, textual features or a combination thereof. Furthermore the inclusion of available meta-data, such as time of broadcast, is shown to lead to very high performance. Latent Dirichlet Allocation is used to model both acoustics and text, yielding fixed dimensional representations of media recordings that can then be used in Support Vector Machines based classification. Experiments are conducted on more than 1200 hours of TV broadcasts from the British Broadcasting Corporation (BBC), where the task is to categorise the broadcasts into 8 genres or 133 show identities. On a 200-hour test set, accuracies of 98.6% and 85.7% were achieved for genre and show identification respectively, using a combination of acoustic and textual features with meta-data.

show abstract

“…It has been found that feature-based adaptation of RNNLMs by augmenting the input with domain-specific auxiliary features provide significant improvements in both perplexity (PPL) and word error rate (WER) [8,2,9,10,4,6,11]. Such features, however, can also include acoustic embeddings [12,13] derived from audio, which might be available for only a subset of the text data, such as the matched in-domain data used for finetuning. In such cases, semi-supervised adaptation approaches can be used.…”

Section: Introductionmentioning

confidence: 99%

Semi-Supervised Adaptation of RNNLMs by Fine-Tuning with Domain-Specific Auxiliary Features

Deena

Madhyastha

et al. 2017

Interspeech 2017

Self Cite

View full text Add to dashboard Cite

Recurrent neural network language models (RNNLMs) can be augmented with auxiliary features, which can provide an extra modality on top of the words. It has been found that RNNLMs perform best when trained on a large corpus of generic text and then fine-tuned on text corresponding to the sub-domain for which it is to be applied. However, in many cases the auxiliary features are available for the sub-domain text but not for the generic text. In such cases, semi-supervised techniques can be used to infer such features for the generic text data such that the RNNLM can be trained and then fine-tuned on the available in-domain data with corresponding auxiliary features.In this paper, several novel approaches are investigated for dealing with the semi-supervised adaptation of RNNLMs with auxiliary features as input. These approaches include: using zero features during training to mask the weights of the feature sub-network; adding the feature sub-network only at the time of fine-tuning; deriving the features using a parametric model and; back-propagating to infer the features on the generic text. These approaches are investigated and results are reported both in terms of PPL and WER on a multi-genre broadcast ASR task.

show abstract

Latent Dirichlet Allocation based organisation of broadcast media archives for deep neural network adaptation

Cited by 10 publications

References 28 publications

Latent Dirichlet Allocation Based Acoustic Data Selection for Automatic Speech Recognition

Latent Dirichlet Allocation Based Acoustic Data Selection for Automatic Speech Recognition

Automatic Genre and Show Identification of Broadcast Media

Semi-Supervised Adaptation of RNNLMs by Fine-Tuning with Domain-Specific Auxiliary Features

Contact Info

Product

Resources

About