Speech recognition by machines and humans

By Richard P. Lippmann. 1997

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Speech & Audio

Citations in the corpus

Achieving Human Parity in Conversational Speech Recognition (2016, background): “In the past, human performance on this task has been widely cited as being 4% [19].”
English Conversational Telephone Speech Recognition by Humans and Machines (2017, background): “When compared with [2] which quotes a WER of 4%, the 5.”
Hierarchical learning : theory with applications in speech and vision (2009, background): “Although MFCCs have become a mainstay of ASR systems, machines still significantly under-perform humans in both noise-free and noisy conditions [68].”
POMDP-Based Statistical Spoken Dialog Systems: A Review (2013, background): “Despite steady progress over the last few decades in speech recognition technology, the process of converting conversational speech into words still incurs word error rates in the range 15%–30% in many real-world operating environments such as in public spaces and in motor cars [7], [8].”
The “Cocktail Party Problem”: What Is It? How Can It Be Solved? And Why Should Animal Behaviorists Study It? (2008, background): “Compared to relatively quiet conditions, however, artificial speech recognition systems exhibit drastically reduced performance in noisy situations with competing speech signals (Barker, 2006; Lippmann, 1997).”
A Tandem Algorithm for Pitch Estimation and Voiced Speech Segregation (2010, background): “For example, automatic speech recognition (ASR) systems perform substantially worse in the presence of interfering sounds [27], [36] and could greatly beneﬁt from an effective speech segregation system.”
Audio-Visual Speech Modeling for Continuous Speech Recognition (2000, background): “Automatic speech recognition (ASR) has been an active research area for several decades, but in spite of the enormous efforts, the performance of current ASR systems is far from the performance achieved by humans: error rates are often one order of magnitude apart [7].”
The IBM 2015 English conversational telephone speech recognition system (2015, background): “To set the baseline, the human word error rate on this task is estimated to be around 4% [2].”
Robust automatic speech recognition with missing and unreliable acoustic data (2001, background): “Lippmann (1997) presents evidence from a wide range of speech recognition tasks which indicates that human error rates are an order of magnitude smaller than those obtained by ASR algorithms for clean speech, and two orders of magnitude smaller for typical noise conditions.”
Automatic speech recognition and speech variability: A review (2007, methodology): “A few years ago, a publication [168] gathered results from bo th human and machine speech recognition, with the goal of stimu lating the discussion on research directions and contribut ing to the understanding of what has still to be done to reach closetohuman performance.”
Audio-Visual Automatic Speech Recognition: An Overview (2004, background): “Indeed, even in “clean” acoustic environments, and for a variety of tasks, state of the art ASR system performance lags human speech perception by up to an order of magnitude (Lippmann, 1997).”
MMAU: A Massive Multi-Task Audio Understanding and Reasoning Benchmark (2024, unknown): “Tasks such as image and speech recognition, for instance, do not demand the expertise of skilled humans and can often be performed by young children (Lippmann, 1997; Gerhardstein & Rovee-Collier, 2002).”
Comparing Human and Machine Errors in Conversational Speech Transcription (2017, background): “A widely cited figure is 4% word error rate (WER), based on [22].”
The INTERSPEECH 2009 emotion challenge (2009, background): “In ASR, a rough estimate for the difference between read and spontaneous data was that, at least to start with, one could expect an error rate for spontaneous data twice the size than the one for read data [10].”
Recent advances in the automatic recognition of audiovisual speech (2003, background): “Indeed, even in “clean” acoustic environments, state of the art ASR system performance lags human speech perception by up to an order of magnitude [1].”