Adrian Mos

Foundation Models shift the interest to adapting models instead of creating proprietary models from scratch. Despite this change, performing hyperparameter optimization (HPO) is still needed. Users adapting systems powered by those models on proprietary data should not considerably increase the overall resource footprint with extensive hyperparameter search. Given that this footprint is also proportional to the data used in HPO, we aim to investigate how a user can effectively reduce the amount of data used, leveraging the deep learning model's empirical facility to output the expected correct result for an item in the dataset. In this work, we describe a methodology for accomplishing this data reduction through estimating a measure of an item's difficulty. This method allows keeping only a portion of data that conserves the overall proportions of item difficulty throughout the dataset while helping order them meaningfully. The rationale is derived from results from curriculum learning research as we try to answer if the adapted models could help organize and select subsets of data representative of the whole. Preliminary results of evaluating the method are provided for image recognition and scientific name entity recognition (NER). We observe that the amount of data for HPO can be reduced as far as 60% and still point to the same choice of hyperparameters compared to using the whole training set.

Mobile applications play an important role in the economy today and there is an increasing trend for app enablement on multiple platforms. However, creating, distributing, and maintaining an application remain expert tasks. Even for software developers, the process can be error-prone and resource-consuming, especially when targeting different platforms simultaneously. Researchers have proposed several frameworks to facilitate cross-platform app development, but little attention has been paid to non-technical users. In this paper, we described the Flow framework, which takes the advantage of domain-specific languages to enable no-code specification for app modeling. The cloud-native coordination mechanism further supports non-technical users to execute, monitor, and maintain apps for any target platforms. User evaluations were conducted to assess the usability and user experience with the system. The results indicated that users can develop apps in Flow with ease, but the prototype could be optimized to reduce learning time and workload.

Machine learning (ML) systems based on deep neural networks are more present than ever in software solutions for numerous industries. Their inner workings relying on models learning with data are as helpful as they are mysterious for non-expert people. There is an increasing need to make the design and development of those solutions accessible to a more general public while at the same time making them easier to explore. In this paper, to address this need, we discuss a proposition of a new assisted approach, centered on the downstream task to be performed, for helping practitioners to start using and applying Deep Learning (DL) techniques. This proposal, supported by an initial testbed UI prototype, uses an externalized form of knowledge, where JSON files compile different pipeline metadata information with their respective related artifacts (e.g., model code, the dataset to be loaded, good hyperparameter choices) that are presented as the user interacts with a conversational agent to suggest candidate solutions for a given task.

Creating and maintaining a modern, heterogeneous set of client applications remains an obstacle for many businesses and individu-als. While simple domain-specific graphical languages and libraries can empower a variety of users to create application behaviors and logic, using these languages to produce and maintain a set of heterogeneous client applications is a challenge. Primarily because each client typically requires the developers to both understand and embed the domain-specific logic. This is because application logic must be encoded to some extent in both the server and client sides. In this paper, we propose an alternative approach, which allows the specification of application logic to reside solely on the cloud. We have built a system where reusable application components can be assembled on the cloud in different logical chains and the client is largely decoupled from this logic and is solely concerned with how data is displayed and gathered from users of the application. In this way, the chaining of requests and responses is done by the cloud and the client side has no knowledge of the application logic. This means that the experts in the domain can build these modular cloud components, arrange them in various logical chains, generate a simple user interface to test the application, and later leave it to client-side developers to customize the presentation and gathering of data types to and from the user. An additional effect of our approach is that the client side developer is able to immediately see any changes they make, while executing the logic residing on the cloud. This further allows more novice programmers to perform these customizations, as they do not need to ‘get the full application working’ and are able to see the results of their code as they go, thereby lowering the obstacles to businesses and individuals to produce and maintain applications. Furthermore, this decoupling enables the quick generation and customization of a variety of application clients, ranging from web to mobile devices and personal assistants, while customizing one or more as needed. Creating new computer applications, distributing them to users, and maintaining them, is complex and time-consuming. Building even a simple application requires years of study. Furthermore, given the number of devices we have today (from smartphones to voice assistants), the task is further complicated and difficult even for experienced programmers. The world is undergoing a digital transformation, and computer applications are now part of our everyday existence and building them is a requirement for many organizations and businesses. How-ever, only companies and individuals with large budgets and skills can build these applications, a consequence of which is an increased digital gap between those with means and those without. We see a clear need to lower the barriers to building these applications so a more diverse set of voices are able to participate and shape our digital futures, as underrepresented groups are often left out of designing and building these platforms. Previous researchers have proposed approaches where applications can be easily specified by people without (or with few) technical skills. However, handling different device platforms and maintenance/distribution were neglected in these efforts. In this paper we propose an approach where the program is executed on the cloud, while the client only needs to display and gather data from users. A result of our approach is that generic front-end applications adapt to any changes to the program being executed on the cloud. Our programs work with end-user apps on any device, the application that the user sees automatically adapts to any changes.

Simple domain-specific graphical languages and libraries can empower a variety of users to create application behavior and logic. However, it remains challenging to produce and maintain a heterogeneous set of client applications based on these descriptions, as each client typically requires the developers to both understand and embed the domain-specific logic. This is because application logic must be encoded to some extent in both the server and client sides.In this paper, we propose an alternative approach, which allows the specification of application logic to reside solely on the cloud. In our system, reusable application components are assembled on the cloud in different logical chains and the client is solely concerned with how data is displayed and gathered from users. In this way, the chaining of requests and responses is done by the cloud and the client side has no knowledge of the application logic. This means that the experts in the domain build modular cloud components, arrange them in logical chains, generate a simple user interface, and later leave it to client-side developers to customize the presentation.

In this paper we present the modeling support infrastructure for domain-specific application definition. This consists of a set of meta-models and the associated generators to allow the definition of reusable and domain-specific behavior blocks, which can later be used to compose complex behaviors. In addition we also present the related visual languages that facilitate the creation of these models.

Purpose: Domain-specific process modeling has been proposed in the literature as a solution to several problems in business process management. The problems arise when using only the generic Business Process Model and Notation (BPMN) standard for modeling. This language includes domain ambiguity and difficult long-term model evolution. Domain-specific modeling involves developing concept definitions, domain-specific processes and eventually industry-standard BPMN models. This entails a multi-layered modeling approach, where any of these artifacts can be modified by various stakeholders and changes done by one person may influence models used by others. There is therefore a need for tool support to keep track of changes done and their potential impacts. The paper aims to discuss these issues. Design/methodology/approach: The authors use a multi-context systems-based approach to infer the impacts that changes may cause in the models; and alsothe authors incrementally map components of business process models to ontologies. Findings: Advantages of the framework include: identifying conflicts/inconsistencies across different business modeling layers; expressing rich information on the relations between two layers; calculating the impact of changes taking place in one layer to the rest of the layers; and selecting incrementally the most appropriate semantic models on which the transformations can be based. Research limitations/implications: The authors consider this work as one of the foundational bricks that will enable further advances toward the governance of multi-layer business process modeling systems. Extensive usability tests would enable to further confirm the findings of the paper. Practical implications: The approach described here should improve the maintainability, reuse and clarity of business process models and in extension improve data governance in large organizations. The approaches described here should improve the maintainability, reuse and clarity of business process models. This can improve data governance in large organizations and for large collections of processes by aiding various stakeholders to understand problems with process evolutions, changes and inconsistencies with business goals. Originality/value: This paper fulfills an identified gap to enabling semantically aided domain–specific process modeling.

Monitoring the execution of business processes and activities composing them is an essential capability of Business Process Management (BPM) Suites. Human tasks are a particular type of business activities, and the understanding of their execution is essential in effectively managing both the processes and human resources. This paper proposes a monitoring framework with a capability to monitor and analyze the human tasks in a domain specific setting and contextually correlate the task execution patterns to the workload distribution on human users. The framework uses the notion of concept probes that match the business concepts used in definition of business processes. The proposed human task monitoring and contextual analysis (HTMCA) component considers multiple artifacts involved in the execution of a human task, rather than focusing only on classic activity/task metrics retrieved from BPM engines.This approach aspires to provide two main advantages to organizations using it. Firstly, it enhances the understanding of the workload of human users that participate in people-intensive business processes under various roles. Secondly, it gives organizations tools and insight for fine-tuning their user performance taking into account the specific context of their business various artifacts domains. The proposed framework builds on previous work that lays the basis of vendor-independent, concept-centric BPM monitoring, and provides the critical missing element of human task understanding. This has the potential to significantly benefit any BPM deployment and the validation work is in advanced stages of building a full prototype that demonstrates this value in a realistic industrial setting.