iROVER: Improving System Combination with Classification

We present an improved system combination technique, iROVER, Our approach obtains significant improvements over ROVER, and is consistently better across varying numbers of component systems. A classifier is trained on features from the system lattices, and selects the final word hypothesis by learning cues to choose the system that is most likely to be correct at each word location. This approach achieves the best result published to date on the TC-STAR 2006 English speech recognition evaluation set.

• Publication year

  2007

• Venue

  North American Chapter of the Association for Computational Linguistics

• Publication date

  2007-04-22

• Fields of study

  Computer Science

• Identifiers
  DOI 10.3115/1614108.1614125
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• iROVER achieves the best result published to date on the TC-STAR 2006 English speech recognition evaluation set.

  Confidence 0.94

  박진우 (dztg5apj7m) extraction뀨 (7c402c1b98) reviewB (s683577b42) review
• A classifier trained on features from the system lattices selects the final word hypothesis by learning cues for choosing the system most likely to be correct at each word location.

  Confidence 0.95

  박진우 (dztg5apj7m) extraction뀨 (7c402c1b98) reviewB (s683577b42) review
• iROVER improves over ROVER and remains consistently better across different numbers of component systems.

  Confidence 0.96

  박진우 (dztg5apj7m) extraction뀨 (7c402c1b98) reviewB (s683577b42) review

• classifier

  model

  A learned model that uses lattice-derived features to decide which system's word hypothesis to select.

  박진우 (dztg5apj7m) extraction뀨 (7c402c1b98) reviewB (s683577b42) review
• component systems

  system

  The individual speech recognition systems whose outputs are combined by the method.

  Aliases: systems

  박진우 (dztg5apj7m) extraction뀨 (7c402c1b98) reviewB (s683577b42) review
• final word hypothesis

  output

  The selected word output at a particular word location in the combined recognition result.

  Aliases: word hypothesis

  박진우 (dztg5apj7m) extraction뀨 (7c402c1b98) reviewB (s683577b42) review
• irover

  method

  An improved speech recognition system combination technique that uses classification to choose word hypotheses from multiple systems.

  Aliases: improved system combination technique

  박진우 (dztg5apj7m) extraction뀨 (7c402c1b98) reviewB (s683577b42) review
• rover

  baseline, method

  A baseline speech recognition system combination method used for comparison with iROVER.

  박진우 (dztg5apj7m) extraction뀨 (7c402c1b98) reviewB (s683577b42) review
• system lattices

  data representation, feature source

  Lattice structures produced by the component recognition systems and used to derive features for classification.

  Aliases: lattices

  박진우 (dztg5apj7m) extraction뀨 (7c402c1b98) reviewB (s683577b42) review
• tc-star 2006 english speech recognition evaluation set

  evaluation setting, benchmark

  The evaluation benchmark on which recognition performance is reported in the abstract.

  Aliases: TC-STAR 2006

  박진우 (dztg5apj7m) extraction뀨 (7c402c1b98) reviewB (s683577b42) review

• Cross-Site and Intra-Site ASR System Combination: Comparisons on Lattice and 1-Best Methods

  2007cited by this paper
• Investigations of issues for using multiple acoustic models to improve continuous speech recognition

  2006cited by this paper
• Frame based system combination and a comparison with weighted ROVER and CNC

  2006cited by this paper
• Improved confusion network algorithm and shortest path search from word lattice

  2005cited by this paper
• Explicit word error minimization using word hypothesis posterior probabilities

  2001cited by this paper
• THE SRI MARCH 2000 HUB-5 CONVERSATIONAL SPEECH TRANSCRIPTION SYSTEM

  2000cited by this paper
• BoosTexter: A Boosting-based System for Text Categorization

  2000influential reference
• Finding consensus in speech recognition: word error minimization and other applications of confusion networks

  2000cited by this paper
• Posterior probability decoding, confidence estimation and system combination

  2000influential reference
• Improved ROVER using Language Model Information

  2000cited by this paper
• Using word probabilities as confidence measures

  1998cited by this paper
• A post-processing system to yield reduced word error rates: Recognizer Output Voting Error Reduction (ROVER)

  1997cited by this paper

• Investigations on Speech Recognition Systems for Low-Resource Dialectal Arabic-English Code-Switching Speech

  2021cites this paper
• Automatic sign language recognition: from video corpora to gloss sentences

  2020cites this paper
• Hybrid Arbitration Using Raw ASR String and NLU Information - Taking the Best of Both Embedded World and Cloud World

  2019cites this paper
• In-Vehicle Voice Interface with Improved Utterance Classification Accuracy Using Off-the-Shelf Cloud Speech Recognizer

  2018cites this paper
• Handwritten Text Recognition Error Rate Reduction in Historical Documents using Naive Transcribers

  2018cites this paper
• Automatic quality estimation for ASR system combination

  2017cites this paper
• Selection and combination of hypotheses for dialectal speech recognition

  2016cites this paper
• Data selection from multiple ASR systems' hypotheses for unsupervised acoustic model training

  2016cites this paper
• Information Fusion Approaches for Distant Speech Recognition in a Multi-microphone Setting

  2016cites this paper
• Semi-Supervised Acoustic Model Training by Discriminative Data Selection From Multiple ASR Systems’ Hypotheses

  2016cites this paper
• Ensemble Learning Approaches in Speech Recognition

  2015cites this paper
• Driving ROVER with Segment-based ASR Quality Estimation

  2015influential citation
• Local System Voting Feature for Machine Translation System Combination

  2015cites this paper
• Confidence-features and confidence-scores for ASR applications in arbitration and DNN speaker adaptation

  2015cites this paper
• A SYSTEM COMBINATION FOR MALAY BROADCAST NEWS TRANSCRIPTION

  2015cites this paper
• A grapheme and phone rescoring combination system for Malay broadcast news recognition

  2015cites this paper
• Speech and Audio Processing for Coding, Enhancement and Recognition

  2014influential citation
• Sub-word based language modeling of morphologically rich languages for LVCSR

  2014cites this paper
• Neural network based feature extraction for speech and image recognition

  2014cites this paper
• Analysis and Combination of Forward and Backward Based Decoders for Improved Speech Transcription

  2013cites this paper
• Dynamic Combination of Automatic Speech Recognition Systems by Driven Decoding

  2013cites this paper
• Combining forward-based and backward-based decoders for improved speech recognition performance

  2013cites this paper
• Attelage de systèmes de transcription automatique de la parole

  2012cites this paper
• The RWTH System Combination System for WMT 2011

  2011influential citation
• DFKI System Combination with Sentence Ranking at ML4HMT-2011

  2011cites this paper
• cROVER: Context-augmented Speech Recognizer based on Multi-Decoders' Output

  2011cites this paper
• Handbook of Natural Language Processing and Machine Translation: DARPA Global Autonomous Language Exploitation

  2011cites this paper
• présentée à l'Université d'Avignon et des Pays de Vaucluse pour obtenir le diplôme de DOCTORAT

  2011cites this paper
• Bayes risk decoding and its application to system combination

  2011cites this paper
• Vers une modélisation statistique multi-niveau du langage, application aux langues peu dotées. (Toward a multi-level statistical language modeling for under-resourced language)

  2010cites this paper
• Systems combination in large vocabulary continuous speech recognition

  2010cites this paper
• Speaker normalisation for large vocabulary multiparty conversational speech recognition

  2009influential citation
• Combined speech decoders output for phoneme recognition enhancement

  2009cites this paper
• Study of algorithms to combine multiple automatic speech recognition (ASR) system outputs

  2009cites this paper
• Combining Spectral Representations for Large-Vocabulary Continuous Speech Recognition

  2008cites this paper
• Complexity of Finding the BLEU-optimal Hypothesis in a Confusion Network

  2008cites this paper
• iCNC and iROVER: the limits of improving system combination with classification?

  2008influential citation
• Automatic sentence structure annotation for spoken language processing

  2008cites this paper