Data networking at BBN

BBN has an illustrious history of contributions to data networking, and has made repeated contributions in the development of networking protocols, network operations, router design, and wireless and satellite networks. Because so much of BBN's early Arpanet work has been documented elsewhere, this history focuses on post-Arpanet contributions that led to the Internet. Arpanet was a continuing source of inspiration, frustration, and innovation, both as a stand-alone network and then as the core of the Internet.     

Speech processing at BBN

This survey of speech processing activities covers a period that began around 1971. Areas of importance - technical as well as historical - include speech recognition and understanding, speech coding, speaker recognition, and speech modification. A number of today's best-regarded techniques in speech and language processing stem from BBN's early work.     

Educational technology at BBN

In exploring ways to apply technology for the purpose of improving learning, we at BBN have conducted basic research in human cognition and learning. We have developed innovative software facilities to extend and enhance mathematics, science, and language learning, and provided professional development workshops to help educators effectively use these new technologies. We have helped schools employ networking facilities to connect teachers and students with each other and with national and global resources. To support student work in mathematics, statistics, physics, and biology, we have developed interactive visual modeling facilities. Finally, we have made major advances in many areas of technical training.     

Control systems R&D at BBN

The primary focus in this article is on BBN's theoretical and applied work on problems involving humans in the information processing and control loop of control systems. This article discusses the evolution and outcome of key technical developments, people involved, and the role of BBN's environment -computational, organizational, and human. The history of the long and productive R&D program from its early beginnings, circa 1960, to the end of the 1980s is the main focus of this article. The history includes efforts in developing decision aids for pilots and for advanced command and control applications, because these can be viewed as natural evolutions of earlier control work. In something of a departure from this main focus, multitarget tracking, which emerged from the control efforts, is discussed because it eventually resulted in major programs for BBN and illustrates how BBN's diverse research efforts and its environment lead to unexpected rewards.     

Medical application of computers at BBN

Early work at BBN involved hospital computer systems, computer aids for the physician's office, data management tools for clinical research, and database and computational support for biomedical research. The work included both development of prototype systems and later deployment of commercially viable software and services. This history also notes some of the challenges of working within an R & D defense-contractor environment and then concludes with lessons learned in developing medical computer applications.     

History of technology transfer at BBN

BBN's basic business since its founding has been contract consulting, research, and development. This article describes BBN's activities from 1948-1989 to transfer technology and intellectual property from its basic sponsored consulting, research, and development business into a variety of commercial and other products and services. This article was intended to give a sense of the varied technology transfer activities that took place at BBN over the better part of 40 years. Although the push to commercialize technology was certainly driven by BBN's board of directors and corporate management, many members of the technical staff also considered it extremely important that their ideas and invention be brought to market. Further, key members of the management and technical leadership teams that compromised BBN's commercial subsidiaries were often drawn from the company's professional services divisions. Finally, management always paid great attention to share in any financial benefit derived from commercial applications of their work.     

Finite element analysis on the BBN butterfly multiprocessor

Applications of the finite element method can easily exhaust CPU and memory resources on conventional uniprocessor machines, and are thus prime candidates for multiprocessor application. In order to investigate various parallel processing methodologies and to evaluate their performance, we have developed a finite element program for the Butterfly Parallel Processor. The overall organization of the program and its data structures are described along with parallel implementations of well-known algorithms for transient and static analysis. Results are presented for two implementations of explicit time integration, and for static analysis using a parallel version of Gaussian elimination. For comparison, results from running the same program on a DEC VAX 11/785 are also included. The transient cases yielded near-linear (within 2.5%) speedup in the full range of 1–64 processors; the static case results showed significant nonlinearity over the same range.     

基于纯度和BBN算法的无监督的话者聚类

由于说话人之间声学特征上的差异,可以将来自于不同说话人的语音段按照话者之间的相似程度进行聚类.在语音段长度不等和说话人数目未知的情况下,本文提出了一种基于纯度信息的不同话者语音段的无监督聚类方法.首先为每个语音段分别建立GMM模型,通过模型似然比计算和GLR距离测度获得语音段之间的相似程度,对语音数据集进行无监督的分级聚类.采用纯度信息和BBN算法进行每级聚类的聚类质量和效果计算,并以此进行分级聚类的剪枝选择.文中给出了平行剪枝和最优剪枝两种剪枝方法及剪枝的评估方法.实验表明,该方法具有较好的聚类效果.     

基于BBN情感模型的和谐人机交互研究

提出了一种将个性、情感、情绪分层表示的思想,并且利用贝叶斯网络进行情感建模,通过虚拟人脸的面部表情来反映情绪的变化,旨在赋予机器类人的情感,达到更加真实和谐的人机交互。最后将此情感模型应用于情感虚拟人交互系统,实验证明,该模型简单、稳定,且易于实现。     

BBN曲面的形状分析与控制

研究了调节凸Bézier曲面、B-样条曲面及NURBS曲面（BBN曲面）一个控制点以后,曲面形状变化的规律.通过将BBN曲面分解成一张凸曲面与具有特殊形状曲面的叠加,建立了曲面变形前后一些几何量与变形位移量之间的数量关系,得到了凸BBN曲面失去凸性的充分条件和判据.相应的结果可应用于调节与控制BBN曲面形状的算法设计.     

Leading a top-notch R&D group in the BBN environment

The author notes his MIT background and transition from MIT Lincoln Laboratory to Bolt Beranek and Newman. He sketches the Arpanet project at BBN from his position as project leader, and he describes BBN's unusual mix of government-funded R&D and commercial activity, including issues and anecdotes involving government contracting, overhead rates, and employee motivation.     

BBN TransTalk: Robust multilingual two-way speech-to-speech translation for mobile platforms

In this paper we present a speech-to-speech (S2S) translation system called the BBN TransTalk that enables two-way communication between speakers of English and speakers who do not understand or speak English. The BBN TransTalk has been configured for several languages including Iraqi Arabic, Pashto, Dari, Farsi, Malay, Indonesian, and Levantine Arabic. We describe the key components of our system: automatic speech recognition (ASR), machine translation (MT), text-to-speech (TTS), dialog manager, and the user interface (UI). In addition, we present novel techniques for overcoming specific challenges in developing high-performing S2S systems. For ASR, we present techniques for dealing with lack of pronunciation and linguistic resources and effective modeling of ambiguity in pronunciations of words in these languages. For MT, we describe techniques for dealing with data sparsity as well as modeling context. We also present and compare different user confirmation techniques for detecting errors that can cause the dialog to drift or stall.     

Performance of a particle-in-cell plasma simulation code on the BBN TC2000

The BBN TC2000 is a multiple instruction, multiple data (MIMD) machine that combines a physically distributed memory with a logically shared memory programming environment using the unique Butterfly switch. Particle-in-cell (PIC) plasma simulations model the interaction of charged particles with electric and magnetic fields. This paper describes the implementation of both a 1-D electrostatic and a 2 1/2-D electromagnetic PIC (particle-in-cell) plasma simulation code on a BBN TC2000. Performance is compared to implementation of the same code on the shared memory Sequent Balance 21000.     

The ABC's of BBN: from acoustics to behavioral sciences to computers

The discipline of psychology, and specifically the concept of man-machine integration, served to organize computer research and development at BBN. This article gives a unifying perspective on the history of computer research and development at Bolt Beranek and Newman Inc. (BBN). It suggests that the firm's original focus on A (acoustics) led to its work in B (behavioral sciences, principally psychology), which in turn led to C (its computer activities) $the three areas then coexisting. In particular, this article suggests that psychological concepts have shaped the company's work on computers from the beginning. The BBN principals desired a larger range of psychoacoustics and a contribution from psychologists (behavioral scientists) and to establish an activity in man-machine integration, as a central thrust in the area of human factors of engineering psychology.     

AlterBBN: A program for calculating the BBN abundances of the elements in alternative cosmologies

We describe AlterBBN, a public C program for evaluating the abundances of the elements generated by Big-Bang Nucleosynthesis (BBN). This program enables the user to compute the abundances of the elements in the Standard Model of cosmology, and additionally provides possibilities to alter the assumptions of the cosmological model in order to study their consequences on the abundances of the elements. In particular the baryon-to-photon ratio, the effective number of neutrinos and the neutron lifetime, as well as the expansion rate and the entropy content of the Universe during BBN can be modified in AlterBBN. Such features allow the user to test the cosmological models by confronting them to BBN constraints. A presentation of the physics of BBN and the features of AlterBBN is provided here under the form of a manual.Program summaryProgram title: AlterBBNCatalogue identifier: AEMH_v1_0Program summary URL: http://cpc.cs.qub.ac.uk/summaries/AEMH_v1_0.htmlProgram obtainable from: CPC Program Library, Queen?s University, Belfast, N. IrelandLicensing provisions: Gnu General Public License version 3No. of lines in distributed program, including test data, etc.: 4611No. of bytes in distributed program, including test data, etc.: 44 657Distribution format: tar.gzProgramming language: C (C99 Standard compliant)Computer: 32- or 64-bit PC, MacOperating system: Linux, MacOSRAM: 100 MbClassification: 1.2, 1.9Nature of problem: Calculation of the abundances of the elements during Big-Bang Nucleosynthesis (BBN).Solution method: AlterBBN computes the abundances of the elements by integrating the BBN differential equations thanks to a second-order Runge–Kutta method, obtaining a matrix equation from implicit differencing and solving this equation using Gaussian elimination.Unusual features: AlterBBN includes the possibility of altering the underlying cosmological model and testing the influence of the cosmological assumptions on the abundances of the elements. Uncertainties can be evaluated by computing the abundances of the elements using lower or higher limits of the nuclear reaction rates.Running time: Compilation: Less than 1 minute. Running: Less than 1 second.     

Analysis of the errors produced by the 2004 BBN speech recognition system in the DARPA EARS evaluations

This paper aims to quantify the main error types the 2004 BBN speech recognition system made in the broadcast news (BN) and conversational telephone speech (CTS) DARPA EARS evaluations. We show that many of the remaining errors occur in clusters rather than isolated, have specific causes, and differ to some extent between the BN and CTS domains. The correctly recognized words are also clustered and are highly correlated with regions where the system produces a single hypothesized choice per word. A statistical analysis of some well-known error causes (out-of-vocabulary words, word fragments, hesitations, and unlikely language constructs) was performed in order to assess their contribution to the overall word error rate (WER). We conclude with a discussion of the lower bound on the WER introduced by the human annotator disagreement.