S. Khudanpur and J. Wu. A maximum entropy language model integrating N-grams and topic dependencies for conversational speech recognition. In Processing, 1999. Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
Knowing What You Don't Know: Roles for Confidence Measures in.. - Williams (1999) (4 citations) (Correct)
....n gram probabilities upon topic and may include part of speech or syntactic constraints, for example. Small reductions in perplexity and WER for a topic constrained ME based language model using substantially fewer parameters than a mixture based n gram is presented for the SWB corpus in [107]. 2 If the words of a language are considered to be emitted from a source, the average degree of uncertainty in the symbol emitted by that source is estimated by H = Gamma 1 L logf P (w1 ; wL )g, where L is assumed to be a sufficiently large sample and P is the probability of the ....
S. Khudanpur and J. Wu. A maximum entropy language model integrating n-grams and topic dependencies for conversational speech recognition. In Proceedings of the International Conference on Acoustics, Speech and Signal Processing, pages 553--556. IEEE, 1999.
Two Decades Of Statistical Language Modeling: Where Do We Go.. - Rosenfeld (2000) (20 citations) (Correct)
....to be included in the model. An automatic iterative procedure for selecting features from a given candidate set is described in [47] An interactive procedure for eliciting candidate sets is described in [53] ME language modeling remains the subject of intensive research; see for example [54, 55, 56, 57, 58]. 3.5. Adaptive models So far we have treated language as a homogeneous source. But in fact natural language is highly heterogeneous, with varying topics, genres and styles. In cross domain adaptation, test data comes from a source to which the language model has not been exposed during ....
Sanjeev Khudanpur and Jun Wu. A maximum entropy language model integrating n-grams and topic dependencies for conversational speech recognition. In Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing, Phoenix, AZ, 1999.
Two Decades Of Statistical Language Modeling: Where Do We Go.. - Rosenfeld (2000) (20 citations) (Correct)
....to be included in the model. An automatic iterative procedure for selecting features from a given candidate set is described in [47] An interactive procedure for eliciting candidate sets is described in [53] ME language modeling remains the subject of intensive research; see for example [54, 55, 56, 57, 58]. 3.5. Adaptive models So far we have treated language as a homogeneous source. But in fact natural language is highly heterogeneous, with varying topics, genres and styles. In cross domain adaptation, test data comes from a source to which the language model has not been exposed during ....
Sanjeev Khudanpur and Jun Wu. A maximum entropy language model integrating n-grams and topic dependencies for conversational speech recognition. In Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing, Phoenix, AZ, 1999.
Maximum Entropy Techniques for Exploiting Syntactic, Semantic.. - Khudanpur, Wu (7 citations) Self-citation (Khudanpur) (Correct)
....topic is trained from all the training data, making it possible to obtain better estimates of the topic independent components of the model. A model with a small number of free parameters follows as a consequence. Section 2. 1 contains a formulation of this model, which was proposed earlier in Khudanpur Wu (1999). Issues in assigning a topic to a test utterance are discussed in Section 2.2. Section 2.3 describes experiments on Switchboard, a corpus of spontaneous American English telephone conversations (see Godfrey et al. 1992) for a description of the corpus) and provides analysis of the results. ....
....and N gram Dependencies Although it is widely acknowledged that the syntactic structure of a sentence should be helpful in predicting words, many challenges must be met to integrate syntactic structure into a language model. An outline of our initial efforts in this direction appears in Wu Khudanpur (1999). The structure of statistical language models and that of syntax are substantially different. Statistical language models are represented as multi dimensional tables of conditional probabilities, while the syntactic structure of a sentence is usually embedded in trees. Therefore, a probabilistic ....
Khudanpur, S. & Wu, J. (1999). A Maximum Entropy Language Model Integrating N-Grams and Topic Dependencies for Conversational Speech Reconition. In Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Phoenix, AZ.
Structural Event Detection for Rich Transcription of Speech - Liu (2004) (Correct)
No context found.
S. Khudanpur and J. Wu. A maximum entropy language model integrating N-grams and topic dependencies for conversational speech recognition. In Processing, 1999.
Dynamic Nonlocal Language Modeling via Hierarchical.. - Florian, Yarowsky (1999) (6 citations) (Correct)
No context found.
Sanjeev Khudanpur and Jun Wu. 1999. A maximum entropy language model integrating n-gram and topic dependencies for conversational speech recognition. In Proceedings of ICASSP.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC