Created within TALIA and implemented within the Social and Creative HP, the tool allows to perform semantic searches, exploring the hidden and unknown connections among MED projects, pilots actions and private&public stakeholders belonging to a same or different thematic domain.
The Semantic Framework has been devised to extract knowledge and strategic information from documents that will support the decision making and the strategic planning, in particular for funded projects analysis. It can be seen as an advanced decision support system able to deal with unstructured documents, which gives to different stakeholders and policy makers an unexpected view of the programme results in order to foster collaboration and innovation.
The Semantic Framework includes the following services:
- Smart Search or Discovery;
- Policy Design Variables;
- Indicator Benchmarking.
SMART SEARCH or DISCOVERY
The Semantic Search Service, also called Smart Search or Discovery allows to perform a search on a specific collection of documents. Differently from a classical search engine, which returns documents when there is an exact matching between terms occurring in the query and those in the documents, semantic search can return documents which are somehow related to the query, i.e., documents which are semantically related with the query.
Semantic technologies behind the smart search service allow to consider the meaning of the words. Using exactly the same words in a query, the semantic search could return different related concepts (and documents) depending on the context.
For example, if we issue the query oil on a collection containing documents related to the topics of Mediterranean diet or green growth, we could obtain as a result a set concepts concerning the olive oil topic, such as olive millers, extra virgin, polyphenols (see Figure 1), while if we issue the same query on a collection of documents related to the topics of renewable energy, we could obtain results mostly related to oil in the sense of combustible, such as diesel gasoline, natural gas, kerosene (see Figure 2).
Given a query expressed as a word or a set of words, the semantic search makes available the following functionalities:
- word similarity: the system can return the set of most similar concepts;
- semantic search: the system can return the set of most related documents;
- document similarity: starting from a document, the system can return the set of the most similar documents.
Semantic technologies also allow to automatically summarize documents, i.e., to automatically produce a shorter version of the document by preserving the meaning and the key contents of the original text. The user can choose the length of the summary in terms of number of sentences or percentage of the original length. This is a very complex task, since it requires to emulate the cognitive capacity of human beings to generate summaries.
POLICY DESIGN VARIABLES
The main objective of this service is to define a bottom-up approach which leverages the MED project outcomes, deliverables and technical reports to unveil meaningful connections useful to the policy making process.
The system allows to create a cross-correlation matrix having on the rows and on the columns concepts which can be autonomously defined by users, by leveraging the knowledge contained in the collection of documents on which the analysis is performed. Each concept is defined using a specific label, and by providing a textual description. For example, we can define the concept Marine renewable energy as “the energy which can be harnessed from the ocean or the marine wind. It is comprised of five main types according to the origin of the extracted power, namely marine wind, surface waves, tides and currents, and thermal and salinity gradients”.
Starting from that description, the system automatically runs the Natural Language Processing pipeline to recognize the concepts occurring in the text and corresponding to those contained in the model generated by the semantic analysis of the documents in the collection. The system is also able to provide a set of most related concepts which could be used to complement the original definition provided by the user. The user can select the concepts to keep.
After the definition of all the concepts on the rows and on the columns, the matrix can be filled in. More specifically, for filling in a specific cell of the matrix the system performs a semantic search using the vector of the column label and the vector of the row label and retrieves a set of most related concepts.
After this process, the matrix is ready to be interpreted by experts, and it is possible to formulate two interpretations:
- one descriptive, based on the assumption that concepts retrieved mirror some evidence coming from the deliverables of the running projects;
- one prescriptive, highlighting the gaps in the previous analysis and the possible action items for the policy maker to reinforce some of the correlations documented in each cell.
This is a concrete way to provide support to the policy maker, starting by a service leveraging the semantic framework platform.
It is a ‘self-diagnostic’ tool assisting MED and EU regions in the development of place-based policies and strategies leveraging creative and social innovation. The service enables benchmarking the performance of MED and EU regions according to the three drivers of Collective Creativity in the Mediterranean that are part of the TALIA Vision and policy learning framework: Community Scale Partnerships, Territorial Innovation and Trans-local Socio-Economic Ecosystems.
Starting from the home page of the system, the system allows to perform several analysis on a specific region.
After clicking on one region, e.g., Apulia, one can retrieve a predefined set of indicators for that region, which are plotted against:
- the average value of that indicator for the MED regions;
- its average value for the “reference regions” (in our example for Southern Italy);
- the national average (in our example for the country of Italy);
- the EU average.
Indicators are provided for three groups of variables, namely:
- contextual variables, including statistics on population, purchasing power standard per inhabitant, as well as the unemployment rate;
- 18 explanatory variables, including design applications, public private co-publications, innovative SMEs collaborating with others, employment in MHT manufacturing and knowledge intensive services, population with tertiary education, median age of population, and so on;
- Creative Innovation index, a synthetic and composite indicator of how a region behaves in relation to Collective Creativity, based on the three elements of the TALIA vision.
Explore the Semantic Search Service [here]
Explore the Policy Design Variables Service [here]
Explore the Indicator Benchmarking Service [here]