Metal clusters deposited on a substrate and positioned at a nanometric distance from a wave-reflecting layer act as nanoresonators able to receive, store and transmit energy within the visible and infrared range of the spectrum. Among the unique effects of these metal nanocluster assemblies are high local field enhancement and nanoscale resonant behaviour driving optical absorption in the visible and infrared range of the spectrum. In these types of devices and sensors the precise nanometric assembly coupling the local field surrounding a cluster is critical for allowing resonance with other elements interacting with this field. In particular, the cluster–mirror distance or the cluster–fluorophore distance gives rise to a variety of enhancement phenomena (e.g. resonant-enhanced fluorescence, REF). Depending on the desired application this ‘resonance’ distance is tuned from 5 up to 500 nm. High-throughput transducers using metal cluster resonance technology are based on surface enhancement of light absorption by metal clusters (surface-enhanced absorption, SEA). These devices can be used for detection of biorecognition binding as well as structural changes in nucleic acids, proteins or any polymer. The optical property made use of in the analytical application of metal cluster films is so-called anomalous absorption. An absorbing film of clusters is positioned 10–400 nm from an electromagnetic wave-reflecting layer. At a well-defined mirror–cluster distance the reflected electromagnetic field has the same phase at the position of the absorbing cluster as the incident field. This feedback mechanism strongly enhances the effective cluster absorption coefficient. These systems are characterized by a narrow reflection minimum whose spectral position shifts sensitively with interlayer thickness, because a given cluster–mirror distance and wavelength defines the optimum phase.

Nanoclusters and nanofilms have the potential to amplify fluorescence and thus to enhance the signal of labeled biomolecules on biochip surfaces. Fluorescent molecules are bound at a certain distance to a resonant layer of a metal or a semiconductor or both, resulting in enhanced absorption and emission of the fluorophore within the electromagnetic near-field. This property makes the system highly useful for interaction studies, including those of DNA and proteins. Due to the amount of data, derived from various sequencing projects and from Proteomic interaction studies within the next years, microarrays (or biochips) will represent a central technology in every lab facilitating high-throughput screening and being easily interfaced with computer databases. However, most chips suffer from the disadvantage of insufficient signal-to-noise (background) ratio and are thus limited to molecules of medium-to-high abundance. Novel approaches are needed for identification of, e.g., low copy RNAs or regulatory proteins. Here we present a study, using novel surface enhanced chips in the standard glass-slide-formats. Applying surface-enhanced fluorescence (SEF), the chips turned out to be useful for interaction studies, such as DNA hybridization, thereby strongly enhancing the on-chip-signals. Compared to standard glass-slide-DNA chips, both the fluorescent signals as well as signal-to-noise ratio were considerably higher.

The most fundamental properties of metal nanoclusters, namely the high local-field enhancement and nanoscale resonance behavior of the cluster electron plasma when exited by electromagnetic radiation, have been used to set up a variety of sensors transducing biorecognitive interactions into optical signals. This paper focuses on applications of resonant-cluster technology, which enabled us to monitor biorecognitive binding of a variety of proteins on a chip, thus constructing high-throughput interaction-screening devices. Decisive for this type of sensor is the nanometric distance from the local field surrounding a cluster to other parts of the sensor interacting with this field. In particular, the cluster-mirror or cluster-fluorophore distance gives rise to a variety of enhancement phenomena. Depending on the desired application this "resonance"- distance is approximately 5-400 nm. All types of sensor can be set up on photolithographically constructed microchips, but microscopic glass slides can also be employed; this also enables the use of standard devices for dotting and read out. Using slide based chips a standard format of 3,200 microdots (125 microm in diameter) was the basis of either microassays applying direct optical transduction via surfaceenhanced absorption or striking for more sensitivity via surface-enhanced fluorescence.

The combination of phage display antibody arrays with a novel nanotransducer technique based on resonant nanoparticles in a nanosandwiched film enables the sensitive parallel screening of proteins. Using the resonance of nanoparticles with their induced mirror dipoles in a thin-film structure, limitations of fluorophores, such as unspecific background and nonvisibility to the eye, can be overcome, thereby leading to an optical signal significantly more sensitive than that of standard colloid techniques. The signal can be both directly observed as a color change of a microdot at the sensor surface and tuned throughout the visible range of the spectrum. Here we report the application of an optical chip using scFv-antibody-antigen interactions. Artificial scFv-antibodies against a variety of proteins, including yeast enzymes and bovine serum albumin (as a standard), were constructed via Phage Display. These scFv-antibodies were then coated onto metal nanoclusters and bound to their antigens that were arrayed as nanodroplets at the resonance layer of the chip. ScFv-Antibody-antigen interaction resulted in a visible array of microdots. Using resonance-enhanced absorption, the absorption signal of the spots was amplified by one to two orders of magnitude (compared to colloid-based techniques). For quantitative analysis, either an 8-micron scanner or a CCD camera (resolution 4 microns) was employed to gain direct-reflection spectra rather than unspecific scatter data (prone to dust and unspecific interaction). Our results demonstrate that this device enables high-throughput proteomics to overcome some limitations of fluorescence, enzyme labels, and colloid techniques.

Automated essay scoring (AES) is gaining increasing attention in the education sector as it significantly reduces the burden of manual scoring and allows ad hoc feedback for learners. Natural language processing based on machine learning has been shown to be particularly suitable for text classification and AES. While many machine-learning approaches for AES still rely on a bag of words (BOW) approach, we consider a transformer-based approach in this paper, compare its performance to a logistic regression model based on the BOW approach, and discuss their differences. The analysis is based on 2088 email responses to a problem-solving task that were manually labeled in terms of politeness. Both transformer models considered in the analysis outperformed without any hyperparameter tuning of the regression-based model. We argue that, for AES tasks such as politeness classification, the transformer-based approach has significant advantages, while a BOW approach suffers from not taking word order into account and reducing the words to their stem. Further, we show how such models can help increase the accuracy of human raters, and we provide a detailed instruction on how to implement transformer-based models for one’s own purposes.