What's the real problem in AEC Innovation?
Architecture, Engineering, and Construction (AEC) companies need to continuously innovate and adopt new technologies in order to stay competitive in the global economy, especially in the wake of the digitalization megatrend. What makes this difficult is partly the diversity of the AEC Industry, which requires competences and knowledge from a wide spectrum of disciplines. The challenges of connecting all relevant stakeholders, as catalyst of innovation, are well-known.
It's difficult for companies to find the best innovation partners, because research and developmental efforts are siloed. This means that, in addition to the barriers of cross-disciplinary engagement (e.g. with ICT), even professors and researchers within the same field are disconnected from the industry and sometimes even one another. This is despite the massive investments that the authorities, such as the European Commission and Finnish Ministries, are devoting to innovation and R&D, including ICT search portals to facilitate matchmaking.
The technical problem is that current keyword- or tag-based search paradigms are not scalable as different systems use different data schemas, thus creating barriers for semantic interoperability. Companies in AEC domain need to readily identify transferable technologies from other domains with which they are not familiar, while different domains use different terminology and descriptions to describe similar things. This makes the searching/finding process very difficult.
How do we approach innovation fragmentation?
We use Artificial Intelligence (AI) and Natural Language Processing to bridge the gap of ontological limitations, while assisting non-expert users explore the relevant innovation call topics through recommendation. Essentially, CICP allows innovation-hungry users to easily/quickly identify most relevant calls for innovation proposals and find the best research partners to collaborate in a team.
The method briefly: we start by mining large amounts of unstructured textual data from research publications, professor profiles, call for proposals etc.. Then, we do algorithmic cleaning and parsing of the raw data before input into a semantic machine learning algorithm. This step allows us to remove a lot of inconsistencies of natural language through “noise reduction”, inferring the links between concepts and topics that are semantically related. Our technique is a modified version of Latent Semantic Analysis, combined with calibrations in the weighting of texts and metadata. In this way, we can train the algorithm to identify transferable technologies that traditionally required experts' eyes i.e. domain-specific conceptual links between topics and texts. We implement the AI as a retrieval- and recommender-engine to address the innovation fragmentation problem through a user interface design.
What is the resulting value of CICP?
It allows non-expert users to quickly and easily find partners across disciplines, as well as relevant innovation opportunities/calls, even if the terminologies are different. Industry can also use the tool to foster closer involvement with educational activities. From the other side, innovation authorities have great challenges with promoting specific call topics to the best candidates. As such, CICP also helps make innovation calls for proposals more visible and reachable, not just to anyone, but to the most relevant audience.
Future vision of CICP
We aim to increase engagement between all innovation agents across disciplines/affiliations, thus accelerating AEC innovation adoption towards dynamic ecosystem beyond the social networks. Our approach can hold promises of scalability and extensibility for the long term.
