Advances in speech transcription at IBM under the DARPA EARS program

This paper describes the technical and system building advances made in IBM's speech recognition technology over the course of the Defense Advanced Research Projects Agency (DARPA) Effective Affordable Reusable Speech-to-Text (EARS) program. At a technical level, these advances include the development of a new form of feature-based minimum phone error training (fMPE), the use of large-scale discriminatively trained full-covariance Gaussian models, the use of septaphone acoustic context in static decoding graphs, and improvements in basic decoding algorithms. At a system building level, the advances include a system architecture based on cross-adaptation and the incorporation of 2100 h of training data in every system component. We present results on English conversational telephony test data from the 2003 and 2004 NIST evaluations. The combination of technical advances and an order of magnitude more training data in 2004 reduced the error rate on the 2003 test set by approximately 21% relative-from 20.4% to 16.1%-over the most accurate system in the 2003 evaluation and produced the most accurate results on the 2004 test sets in every speed category.     

Implementing Information Systems: Some Practical Advice and a Richer Model

Recent research suggests that implementing information systems presents considerable difficulties and that many implementations are total or partial failures. This paper argues that what both practitioners and students require are richer and more acceptable models of information systems implementation. Accordingly, case study data concerning the introduction of manufacturing resource planning (MRP II) are used to illustrate five patterns of behaviour (rites) which capture important social actions; and four components of changing psychological orientation. It is argued that appreciation of the rites and psychological developments identified here will support effective change programmes. Guidelines based on our model are provided for practitioners.     

Robust channel prediction technique for decision directed RAKEreceivers

The authors describe a robust channel prediction technique for a direct sequence spread spectrum (DS-SS) system in a fast fading environment. Coherent diversity reception is employed for improved performance. To estimate the channel response, an FIR type linear prediction filter is employed for each RAKE tap. The stability of the decision directed receiver is achieved through differential encoding of the data bits. It is demonstrated through simulations that the performance of the proposed decision directed receiver is better than a receiver which relies on a pilot signal     

Spread Spectrum PN Code Acquisition Using Hybrid Correlator Architectures

A hybrid correlator architecture is described which combines the serial structure of an active correlator with the parallel structure of a matched filter correlator. The mean PN code acquisition time performance of this hybrid serial-parallel correlator structure is analysed. Results are shown which compare the acquisition performance of the serial, parallel, and serial-parallel structures. The results are for a PN code length of 64 code chips and assumes a Gaussian channel with the receiver detection threshold set to obtain a constant false alarm rate. An enhancement to the serial-parallel acquisition algorithm is also described which can increase the acquisition time performance by about 15% for typical operating conditions. Overall the results demonstrate that the hybrid correlator can provide rapid code acquisition with a limited receiver complexity.     

Validation of a Method for Quantitation of Soybean Lectin in Commercial Varieties

Soybean agglutinin (SBA) protein, also known as soybean lectin, is regarded as an anti‐nutrient due to its negative effect on the ability of monogastric animals to gain weight following consumption of raw soybean seed. Historically, SBA has been measured using a time‐consuming and cumbersome hemagglutination procedure. The objective of our research was to obtain a validated methodology that is precise and accurate in the measurement of SBA while allowing minimally equipped laboratories to effectively conduct the analysis, thus our focus was on evaluating an existing commercially available ELISA, an enzyme‐linked‐lectin‐assay (ELLA), and a hybrid ELISA/ELLA. A new ELLA technique that can detect and quantify lectins was chosen and modified specifically for the quantitation of SBA in soybean seed. The proposed ELLA methodology is similar to a standard sandwich ELISA, and uses polyacrylamide‐linked N‐acetylgalactosamine (Gal–NAc–PAA) for a capture phase and the biotinylated version (Gal–NAc–PAA–Biotin) for detection. Based upon the validation data, the ELLA method can precisely and accurately determine soybean lectin levels in soybean seed. The validated ELLA method was used to quantify SBA in nine commercial soybean varieties introduced between 1972 and 2008 and demonstrated that the natural variability of SBA is subject to the effects of genotype and environment.     

Non‐Covalent Functionalization of Graphene Using Self‐Assembly of Alkane‐Amines

A simple, versatile method for non‐covalent functionalization of graphene based on solution‐phase assembly of alkane‐amine layers is presented. Second‐order Møller–Plesset (MP2) perturbation theory on a cluster model (methylamine on pyrene) yields a binding energy of ≈220 meV for the amine–graphene interaction, which is strong enough to enable formation of a stable aminodecane layer at room temperature. Atomistic molecular dynamics simulations on an assembly of 1‐aminodecane molecules indicate that a self‐assembled monolayer can form, with the alkane chains oriented perpendicular to the graphene basal plane. The calculated monolayer height (≈1.7 nm) is in good agreement with atomic force microscopy data acquired for graphene functionalized with 1‐aminodecane, which yield a continuous layer with mean thickness ≈1.7 nm, albeit with some island defects. Raman data also confirm that self‐assembly of alkane‐amines is a non‐covalent process, i.e., it does not perturb the sp² hybridization of the graphene. Passivation and adsorbate n‐doping of graphene field‐effect devices using 1‐aminodecane, as well as high‐density binding of plasmonic metal nanoparticles and seeded atomic layer deposition of inorganic dielectrics using 1,10‐diaminodecane are also reported.     

From data to knowledge: The interaction between data management systems in educational institutions and the delivery of quality education

No school is an island; it is a part of a continuum or a pipeline of institutions which together form an educational pipeline through which groups of students pass. To turn a body of data into useful information for knowledge-based decision-making at any level, data must be collected, organised, analysed and reflected upon. The purpose of this paper is to discuss how schools and other educational institutions can not only collect better data but learn how to transform that data so that the information held within can be effectively shared among all stakeholders. This process will help to ensure that the school and the entire education system provide a more seamless and effective educational pipeline for students, and ultimately improve the quality of education delivered in the country as a whole.