Jiawei Zhou, Jason Eisner, Michael Newman, Emmanouil Antonios Platanios, Sam Thomson. Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2022.

Universal quantum computation requires the implementation of a logical non-Clifford gate. In this paper, we characterize all stabilizer codes whose code subspaces are preserved under physical T and T†gates. For example, this could enable magic state distillation with non-CSS codes and, thus, provide better parameters than CSS-based protocols. However, among non-degenerate stabilizer codes that support transversal T, we prove that CSS codes are optimal. We also show that triorthogonal codes are, essentially, the only family of CSS codes that realize logical transversal T via physical transversal T. Using our algebraic approach, we reveal new purely-classical coding problems that are intimately related to the realization of logical operations via transversal T. Decreasing monomial codes are also used to construct a code that realizes logical CCZ. Finally, we use Ax's theorem to characterize the logical operation realized on a family of quantum Reed-Muller codes. This result is generalized to finer angle Z-rotations in https://arxiv.org/abs/1910.09333.

We consider the problem of classical strong (amplitude-wise) simulation of n-qubit quantum circuits, and identify a subclass of simulators we call monotone. This subclass encompasses almost all prominent simulation techniques. We prove an unconditional (i.e. without relying on any complexity-theoretic assumptions) and explicit (n - 2)(2n-3- 1) lower bound on the running time of simulators within this subclass. Assuming the Strong Exponential Time Hypothesis (SETH), we further remark that a universal simulator computing any amplitude to precision 2-n/2 must take at least 2n-o(n)time. We then compare strong simulators to existing SAT solvers, and identify the time-complexity below which a strong simulator would improve on state-of-the-art general SAT solving. Finally, we investigate Clifford+T quantum circuits with t T-gates. Using the sparsification lemma, we identify a time complexity lower bound of 22.2451×10-8tbelow which a strong simulator would improve on state-of-the-art 3-SAT solving. This also yields a conditional exponential lower bound on the growth of the stabilizer rank of magic states.

Abstract Information systems (IS) project failure is a costly and common problem. This is despite advances in development tools and technologies. In this article, we argue that one reason for this is the failure of project post‐mortems to generate constructive “lessons learned” from previous projects. Over time, the ineffective practices would persist in the organization, rendering it resistant to change. The attribution theory literature serves as one of the few promising theoretical bases to explain why project post‐mortems fail. A case study of a project post‐mortem undertaken for an abandoned electronic procurement system project is discussed and analyzed. We identify five antecedent conditions of attribution error: the presence of self‐appointed mindsets, the general persistence of negative beliefs, memory decay, selective recall of project events and the influence of power dynamics within the organization. We discuss the research and practical implications of these findings and suggest how the problem of attribution error may be minimized in project post‐mortems.

NIST has introduced two-speaker tasks in its Speaker Recognition evaluation scheme. We discuss Dragon Systems' explorations of these tasks using a GMM-based speaker ID system. We present baseline results submitted for the NIST two-speaker evaluation in April 1999. We also report on some "cheating" experiments and investigations of alternative segmentation and channel normalization techniques, which illuminate some challenges specific to the two-speaker environment.

This paper describes recent changes in Dragon's speech recognition system which have markedly improved performance on conversational telephone speech. Key changes include: the conversion to modified perceptual linear prediction (PLP)-based cepstra from mel-cepstra; the replacement of our usual IMELDA transformation by a new transform using "semi-tied covariance"; a new multi-pass adaptation protocol; probabilities on alternate pronunciations in the lexicon; the addition of word-boundary tags in our acoustic models and the redistribution of model parameters to build fewer output distributions but with more mixture components per model.

In this paper we shall describe key improvements to Dragon’s Broadcast News Transcription System, which include: the addition of a speaker-change detection algorithm to our preprocessing subsystem, a new diagonalizing transformation trained using semi-tied covariances, and the addition of probabilities on pronunciations. This new transcription system yields a word error rate of 15.2% on the 1997 evaluation test data, and 14.5% őn the 1998 evaluation test data.

We present a study of data simulated using acoustic models trained on Switchboard data, and then recognized using various Switchboard-trained acoustic models. When we recognize real Switchboard conversations, simple development models give a word error rate (WER) of about 47 percent. If instead we simu-late the speech data using word transcriptions of the conversation, obtaining the pronunciations for the words from our recognition dictionary, the WER drops by a factor of five to ten. In a third type of experiment, we use human-generated phonetic transcripts to fabricate data that more realistically represents conversational speech, and obtain WERs in the low 40’s, rates that are fairly similar to those seen in actual speech data. Taken as a whole, these and other experiments we describe in the paper suggest that there is a substantial mismatch between real speech and the combination of our acoustic models and the pro-nunciations in our recognition dictionary. The use of simulation appears to be a promising tool in our efforts to understand and reduce the size of this mismatch, and may prove to be a generally valuable diagnostic in speech recognition research. 1.

This paper describes improvements made to Dragon's speech recognition system which have improved performance on Switchboard recognition by roughly 10 percentage points in the past year. These features include the use of rapid speaker adaptation, a move from a 20 to a 10 msec frame rate for recognition, expansion of the acoustic training set and lexicon, and the introduction of interpolated language models. Preliminary results applying this Switchboard-trained system to conversations drawn from the English Call-Home corpus are also quite strong, suggesting that this technology ports well to novel tasks. Finally, the paper includes a report on several research projects which show promise of further reducing the error rate.

We have developed a highly accurate automatic language identification system based on large vocabulary continuous speech recognition (LVCSR). Each test utterance is recognized in a number of languages, and the language ID decision is based on the probability of the output word sequence reported by each recognizer. Recognizers were implemented for this test in English, Japanese, and Spanish, using the Ricardo corpus of telephone monologues. When tested on the OGI corpus of digitally recorded telephone speech, we obtained error rates of 3% or lower on 2-way and 3-way closed-set classification of ten-second and one-minute speech segments.

The stressful nature of computer-based work is often highlighted in the research literature. In this study, we argue that a well designed computer system should realize the twin aims of enhancing performance and lowering stress. This paper reports on a psychophysiological field study of the implementation of a command-and-control system in an ambulance service. The evaluation revealed both improvements in operator performance and a reduction in stress levels. In particular, it was found that computer support reduced both systolic blood pressure and subjective anxiety during conditions of peak workload. These findings are discussed in terms of Turner and Karasek's integrated model of the relationships between computer system design, task performance and well-being. The success of the computer system was attributed to the support that it gave operators; by enhancing their degree of control it enabled them to cope better in a highly demanding work environment. The study shows that psychophysiological techniques have a valuable role to play in system design/evaluation; and more generally, that systems development methodologies should take greater account of applied psychological research, especially in areas such as stress.

We trace the process of developing and implementing a materials management system in one company over a 15-year period. Using a process research model developed by Newman and Robey, we identify 44 events in the process and define them as either encounters or episodes. Encounters are concentrated events, such as meetings and announcements, that separate episodes, which are events of longer duration. By examining the sequence of events over the 15 years of the case, we identify a pattern of repeated failure, followed by success. Our discussion centers on the value of detecting and displaying such patterns and the need for theoretical interpretation of recurring sequences of events. Five alternative theoretical perspectives, originally proposed by Kling, are used to interpret the sequential patterns identified by the model. We conclude that the form of the process model allows researchers who operate from different perspectives to enrich their understanding of the process of system development.

User resistance to information systems development and use has plagued the computing community for decades. In this paper, we survey the literature, summarizing the key concepts associated with resistance. Using a case study, we seek to show how resistance occurs in practice. It is shown that resistance is a complex phenomenon which defies simple prescriptions.