Automatic summarization of voicemail messages using lexical and prosodic features

Abstract

This aticle presents trainable methods for extracting principal content words from voicemail messages. The short text summaries generated are suitable for mobile messaging applications. The system uses a set of classifiers to identify the summary words with each word described by a vector of lexical and prosodic features. We use an ROC-based algorithm, Parcel, to select input features (and classifiers). We have performed a series of objective and subjective evaluations using unseen data from two different speech recognition systems as well as human transcriptions of voicemail speech.

Speaker verification using sequence discriminant support vector machines

“…The formulae for the derivatives when the generative model is an HMM may be found in [6]. Let (23) so that the diagonal covariance GMM likelihood is (24) where is the set of parameters in the GMM, . In particular, is the prior of the th Gaussian component of the GMM, is the mean vector of the th component, and is the corresponding diagonal covariance vector.…”

“…Cepstral mean subtraction was applied to remove the effects of the communication channel. Silence frames within each utterance were segmented out using a multilayer perceptron pre-trained on a different dataset [23].…”

Automatic Detection, Indexing, and Retrieval of Multiple Attributes from Cross‐Lingual Multimedia Data

“…all shouted segments, because shouting may indicate extra importance in some situations. To determine the vocal effort of a speaker, LPC (Linear Predictive Coding) analysis is performed using several different orders (13)(14)(15)(16)(17)(18)(19). The best fit to the incoming spectrum is then determined, and used for inverse-filtering.…”

Automatic acquisition of grammatical types for nouns

“…There has been much significant progress made in speech summarization for English or Japanese text and audio sources (Hori and Furui, 2003;Inoue et al, 2004;Koumpis and Renals, 2005;Maskey and Hirschberg, 2003;Maskey and Hirschberg, 2005). Some research efforts have focused on summarizing Mandarin sources Huang et al, 2005), which are dependent on lexical features.…”

“…Some research efforts have focused on summarizing Mandarin sources Huang et al, 2005), which are dependent on lexical features. Considering the difficulty in obtaining high quality transcriptions, some researchers proposed speech summarization systems with non-lexical features (Inoue et al, 2004;Koumpis and Renals, 2005;Maskey and Hirschberg, 2003;. However, there does not exist any empirical study on speech summarization without lexical features for Mandarin Chinese sources.…”