Project Overview
| Category | Information |
|---|---|
| Acronym | StairwAI |
| Project ID | 101017142 |
| Status | Ended |
| Entity type | University / Research |
| Start Date | 01 January 2021 |
| End Date | 31 December 2023 |
| Last Updated Date | 02 May 2025 |
| Overall Budget | € 5,371,707.18 |
| EU Contribution | € 5,116,631.25 |
| Coordinated by | ALMA MATER STUDIORUM - UNIVERSITA DI BOLOGNA |
| Funded under | Horizon2020 Framework Programme |
CORDIS link
Details here
Objective
The most fundamental need of the AI4EU on-demand Platform at this stage in its development is the introduction of improved functionality to allow for the easy engagements of it core targets stakeholders, namely SMEs. While additional resources and strong domain-specific solutions are also needed, it will be a wasted enterprise without a simple mechanism to match users to these assets. What is needed is a Stairway to AI, a linking bridge between users in a low-tech level to the higher-level AI resources that have the potential to transform both their business.
The StairwAI project targets low-tech users with the goal of facilitating their engagement on the AI on-demand Platform. This will be achieved through a new service layer enriching the functionalities of the on-demand platform and containing:
- a multi-lingual interaction layer enabling conversations with the Platform in the user’s own language,
- a horizontal matchmaking service for the automatic discovery of AI assets (tools, data sets, AI experts, consultants, papers, courses etc.) meeting the user business needs and,
- a vertical matchmaking service that will dimension and provision hardware resources through a proper hardware provider (HPC, Cloud and Edge infrastructures).
These services are designed and implemented by using techniques in an AI for AI fashion. The AI techniques deployed in the development of the services are natural language processing for the multi-lingual interaction, constraint solving, optimization and machine learning for horizontal and vertical matchmaking, knowledge representation for organizing the platform AI assets, reputation and fairness mechanisms to improve the matching results.
StairwAI will have a tremendous impact on the sustainability, collaboration opportunities, accessibility and fairness of the AI on-demand Platform, enabling the definition of proper business models for the uptake of AI bringing new value for EU industry.
Tags
- Natural multi-language processing
- matchmaking
- hardware dimensioning and provisioning
- Low-tech SMEs
Project Report Summary
Context and Overall Objectives
The StairwAI project aims at supporting companies, and SMEs in particular, in their adoption and uptake of AI technology. For this purpose, three services for the low-tech SMEs are developed for providing
- natural-language, multi-lingual interaction,
- AI asset (experts, courses, documents, tools, success stories) discovery, and
- hardware resource dimensioning.
For this aim, the project will use NLP to facilitate its use for companies describing case studies in different languages and will address the solutions using advanced algorithms on matchmaking, ontologies and knowledge graphs to structure the information adequately, and reputation and fairness to improve the quality of results.
Specifics objectives are:
- to lower the barriers to AI accessibility for SMEs and low-tech SMEs;
- to provide efficient mechanisms for matchmaking;
- to provide mechanisms for the dimension of the hardware resources for AI software assets;
- to integrate these services on the European AI-on-demand platform;
- to validate the tools and services through a subset of participants on the open calls
- to structure the information efficiently in a flexible manner to enlarge it if needed under different contributors
Work Performed and Main Achievements
During this period, the requirements and architecture of the StairwAI service have been developed through a series of workshops and surveys to the main stakeholders (SMEs) within WP2.
The design of the architecture has been developed with the contribution of all WP leaders. A knowledge graph using graph databases has been developed (WP3) together with an ontology of terms in cooperation with the Ontologies WG initiated in AI4EU project and is currently active for supporting the AI on-demand platform.
The NLP service is under testing and provides an architecture to automatic label terms from voice or written inputs from the stakeholders using the services (WP4).
The main algorithms for matchmaking and the core of the service are already developed and tested in the internal environment (WP5) and will be made operational soon.
The vertical matchmaking (WP6) engine is already operational and the benchmarks are implemented using the Bonseyes marketplace and platform.
The open calls - the first round – have already been completed and are working, (WP7).
The positioning of the project in the community through the WP8 has been achieved, synchronizing the effort through a WP led by StairwAI with the other ICT49.
In addition, engagement with the NoEs, and the DIHs is currently a work in progress and fruitful through multiple channels.
Dissemination has been done through several publications and the engagement with different stakeholder groups on a series of events, for example, Digital SMEs, ICT-49 clusters, Network of excellence, etc.
Results and Impact
The progress beyond the state of the art is mainly related to original horizontal and vertical matchmaking approaches.
In particular, the state of the art has been advanced on the dimensioning of hardware, as the algorithms have been proven to successfully address the problem.
The expected results are the StairwAI service integrated with the AI on demand to become a key referent in the European AI community.
This service can be split into other 5 key subsystems that can be reused independently:
- NLP,
- vertical matchmaking,
- horizontal matchmaking,
- benchmarking
- and the knowledge and data model.
Other results are aligned with the applicants to StairwAI as the project contributes to low-tech SMEs growing and evolving on their uptake of AI technology systems.
Finally, the impact expected is mainly devoted to
- contribute to the sustainability and the enrichment of the AI-on-demand platform
- accelerate the adoption of AI for SMEs and low-tech SMEs uptake of AI through the connection with AI assets and AI experts.
The use of the service will facilitate personnel upskilling on AI in those companies.
Finally, an improved competition with non-EU providers on cloud and edge computing using European hardware resources through vertical matchmaking is envisioned.
Here are the two sections in a table format without links:
Deliverables
| Deliverable | Type | Publication Date |
|---|---|---|
| Matchmaking algorithms V1 | Demonstrator, pilot, prototype | 06 November2022 |
| Data management plan | Open Research Data Pilot | 06 November2022 |
| First Call Announcement and Guide for Applicant | Document, report | 06 November2022 |
| List of FSTP Beneficiaries -First call | Document, report | 06 November2022 |
| Requirements for the AI-on-demand platform-1st version | Document, report | 06 November2022 |
| Experts and HW Providers Catalogue-1st version | Document, report | 06 November2022 |
| Community Kickoff Workshop | Document, report | 06 November2022 |
| Requirements for the AI-on-demand platform-2nd version | Document, report | 06 November2022 |
| Dissemination and Communication Plan | Document, report | 06 November2022 |
| National and international networking strategy report and outlook beyond the project-1st version | Document, report | 06 November2022 |
| StairwAI web site | Other | 06 November2022 |
| Design of the knowledge representation in the StairwAI AI Asset Management System-1st version | Document, report | 06 November2022 |
| Design of the knowledge representation in the StairwAI AI Asset Management System-2nd version | Document, report | 06 November2022 |
| StairwAI AI Asset Management System-1st version | Demonstrator, pilot, prototype | 06 November2022 |
Publications
| Publication | Year |
|---|---|
| A Mechanism for Reasoning over Defeasible Preferences in Arg2P | 2021 |
| Shallow2Deep: Restraining Neural Networks Opacity through Neural Architecture Search | 2021 |
| Symbolic knowledge extraction from opaque ML predictors in PSyKE: Platform design & experiments | 2022 |
| On the Design of PSyKE: A Platform for Symbolic Knowledge Extraction | 2021 |
| GridEx: An Algorithm for Knowledge Extraction from Black-Box Regressors | 2021 |
| Graph Neural Networks as the Copula Mundi between Logic and Machine Learning: A Roadmap | 2021 |
| HADA: An automated tool for hardware dimensioning of AI applications | 2022 |
| Symbolic Knowledge Extraction from Opaque Machine Learning Predictors: GridREx & PEDRO | 2022 |
| A Privacy-Preserving Dialogue System Based on Argumentation | 2022 |
| Multi-Task Attentive Residual Networks for Argument Mining | 2023 |
| Online AutoML: an adaptive AutoML framework for online learning | 2022 |
| Load Classification: A Case Study for Applying Neural Networks in Hyper-Constrained Embedded Devices | 2021 |
| Discovering Business Processes models expressed as DNF or CNF formulae of Declare constraints | 2022 |
| A Sentiment and Emotion Annotated Dataset for Bitcoin Price Forecasting Based on Reddit Posts | 2022 |
| Detecting Arguments in CJEU Decisions on Fiscal State Aid | 2022 |
| Multimodal Argument Mining: A Case Study in Political Debates | 2022 |
| Argumentation Structure Prediction in CJEU Decisions on Fiscal State Aid | 2023 |
| Disruptive situation detection on public transport through speech emotion recognition | 2024 |
| A Dataset of Argumentative Dialogues on Scientific Papers | 2023 |
| Adding preferences and moral values in an agent-based simulation framework for high-performance computing | 2023 |
| Papr Readr Bot: A Conversational Agent to Read Research Papers | 2022 |
| Constrained Hardware Dimensioning for AI Algorithms | 2022 |
| Discovering Business Processes models expressed as DNF or CNF formulae of Declare constraints | 2022 |
| Shape Your Process: Discovering Declarative Business Processes from Positive and Negative Traces Taking into Account User Preferences | 2022 |
| Encouraging AI Adoption by SMEs: Opportunities and Contributions by the ICT49 Project Cluster | 2023 |
| TinderAI: Support System for Matching AI Algorithms and Embedded Devices | 2023 |
| Combining WordNet and Word Embeddings in Data Augmentation for Legal Texts | 2022 |
| On the Definition of Prescriptive Annotation Guidelines for Language-Agnostic Subjectivity Detection | 2023 |
| Explainable Agents Adapt to Human Behaviour | 2023 |
| TinderAI: Support System for Matching AI Algorithms and Embedded Devices | 2023 |