Open Access

Many assistant systems have evolved toward assistant platforms. These platforms combine a range of resources from various actors via a declarative and generative interface. Among the examples are voice-oriented assistant platforms like Alexa and Siri, as well as text-oriented assistant platforms like ChatGPT and Bard. They have emerged as valuable tools for handling tasks without requiring deeper domain expertise and have received large attention with the present advances in generative artificial intelligence. In view of  their growing popularity, this Fundamental outlines the key characteristics and capabilities that define assistant platforms. The former comprise a multi-platform architecture, a declarative interface, and a multi-platform ecosystem, while the latter include capabilities for composition, integration, prediction, and generativity. Based on this framework, a research agenda is proposed along the capabilities and affordances for assistant platforms.

[Context and motivation] Many of today´s systems use artificial intelligence, where Machine learning (ML) is a subfield. Requirements engineering (RE) addresses the needs of the stakeholders for systems development. In particular, systems with ML components require specific non-functional requirements (NFRs) to define ML relevant details, such as quality aspects of training datasets, retrainability of ML-models or specifics of the ML training pipeline. [Problem] The specific application of RE techniques in practical use to systems with ML components is not yet completely understood. It is not clear, which techniques for elicitation, documentation of requirements can be used efficiently for ML based systems. [Ideas and results] Based on a systematic mapping study; we identify 58 NFRs used in studies to describe particular ML requirements. Through an online survey and expert interviews, we identified 30 NFRs that need to be considered in particular for systems with ML components. For the documentation of the highly relevant NFRs, a template was designed, evaluated and optimized in two IT companies. This template helps to ensure consistent documentation of the NFRs. [Contribution] Based on the systematic mapping study, the online survey and the expert interviews, we provide a list of relevant NFRs and a template for documenting the NFRs for systems with ML components. We validated the proposed template using a real world case in the context of two IT industry companies and several software projects. The evaluation shows an increased completeness of requirements.

Modern persistent Key/Value-Stores operate on updatable datasets — massively exceeding the size of available main memory. Tree-based key/value storage management structures became particularly popular in storage engines. B+-Trees allow constant search performance, however write-heavy workloads yield inefficient write patterns to secondary storage devices and poor performance characteristics. LSM-Trees overcome this issue by horizontal partitioning fractions of data — small enough to fully reside in main memory, but require frequent maintenance to sustain search performance. To this end, firstly, we propose Multi-Version Partitioned BTrees (MV-PBT) as sole storage and index management structure in key-sorted storage engines like Key/Value-Stores. Secondly, we compare MV-PBT against LSM-Trees. The logical horizontal partitioning in MV-PBT allows leveraging recent advances in modern B+-Tree techniques in a small transparent and memory resident portion of the structure. Structural properties sustain steady read performance, even on historical data, and yield efficient write patterns as well as reduced write-amplification. We integrate MV-PBT in the WiredTiger key/value storage engine. MV-PBT offers an up to 2x increased steady throughput in comparison to LSM-Trees and several orders of magnitude in comparison to B+-Trees in a YCSB workload. Moreover, MV-PBT exhibits robust time-travel query performance and outperforms LSM-Trees by 20% and B-Trees by an order of magnitude.

Background: Wideband acoustic immittance (WAI) and wideband tympanometry (WBT) are promising approaches to improve diagnosis accuracy in middle-ear diagnosis, though due to significant interindividual difference, their analysis and interpretation remains challenging. Recent approaches have come up, implementing machine learning (ML) or deep learning classifiers trained with measured WAI or WBT data for the classification of otitis media or otosclerosis. Also, first approaches have been made in identifying important regions from the WBT data, which the classifiers used for their decision-making. Methods: Two classifiers, a convolutional neural network (CNN) and the ML algorithm extreme gradient boosting (XGB), are trained on artificial data obtained with a finite-element ear model providing the middle-ear measurements energy reflectance (ER), pressure reflectance phase, impedance amplitude and phase. The performance of both classifiers is evaluated by cross-validation on artificial test data and by classification of real measurement data from the literature using the metrics macro-recall and macro-F1 score. The feature contributions are quantified using the feature importance ‘gain’ for XGB and deep Taylor decomposition for CNN. Results: In the cross-validation with artificial data, the macro-recall and macro-F1 scores are similar, namely 91.2% for XGB and 94.5% for CNN. For the classification with real measurement data the macro-recall and macro-F1-score were 81.8% and 38.2% (XGB) and 81.0% and 54.8% (CNN), respectively. The key features identified are ER between 600–1,000 Hz together with impedance phase between 600–1,000 Hz for XGB and ER up to 1,500 Hz for CNN. Conclusions: We were able to show that the applied classifiers CNN and XGB trained with simulated data lead to a reasonably well performance on real data. We conclude that using simulation-based WAI data can be a successful strategy for classifier training and that XGB can be applied to WAI data. Furthermore, ML interpretability algorithms are useful to identify relevant key features for differential diagnosis and to increase confidence in classifier decisions. Further evaluation using more measured data, especially for pathological cases, is essential.

Object grasping is a crucial task for robots, inspired by nature, where humans can flexibly grasp any object and detect whether it is slipping from grasp or not, more by the sense of touch than vision. In this work we present a bionic gripper with an Edge-AI device that is able to dexterously grasp the handled objects, sense and predict their slippage. In this paper, a bionic gripper with tactile sensors and a time-of-flight sensor is developed. We propose a LSTM model which is used to detect (incipient) slip/slippage, where a 6 degree-of-freedom robot manipulator is used for data collection and testing. The aim of this paper is to develop an efficient slip detection system which we can deploy on the edge device on our gripper, so it can be a stand-alone product that can be attached to almost any robotic manipulator. We have collected a dataset, trained the model and achieved a slip detection accuracy of 95.34%. Due to the efficiency of our model we were able to implement the slip detection on an edge device. We use the Nvidia Jetson AGX Orin development board to show the inference/prediction in a real-time scenario. We demonstrate in the our experiments how the on-gripper slip detection capability allows more robust grasping as the grip force is adjusted in response to a slippage.