IEEE Technology and Society Magazine - March 2021 - 91

the ontology and, on the other hand, the relationships between said instances [1], [34]. We observe,
in turn, that there are two widely used approaches
to tackle these knowledge extraction tasks: entity
recognition (NER) and the definition of linguistic
patterns. Furthermore, these techniques can also be
used in combination [36].
SIAMES focuses its research in this field on the
sections related to feature extraction through natural language processing (NLP), specifically, through
NER, but combining it with the CBR approach and
integrating both approaches.
In a nutshell, the present work entails an original
investigation both by the application of the aforementioned technologies in a little-explored field, such as
the measurement of social impact for social enterprises, and by their integration, innovative in itself.

Description
The strategy being followed in SIAMES to provide
a set of valid indicators to measure the social impact
of social projects combines principally text mining
and CBR techniques.
As shown in Figure 1, we start from a corpus of
reports (described in the next section) from which
we will extract information to build our case database, modeled as an ontology. SIAMES will be able
to retrieve the most similar cases for any given case
entered by a user, after comparing the cases in its
database making use of a semantic similarity function. Finally, whenever possible, the new case will
be included in the case database.

Description of social impact reports corpus
The initial corpus was made up of 380 measurement reports freely published on the Internet; about
half of them coming from Social Value U.K.5 On
this corpus, a manual selection of the 120 that best
meet our purpose has been made first. Basically, we
selected those that, in our opinion, reflect with sufficient clarity the indicators used, also appreciating
those that also include proxies and/or verification
sources. These include a large number of reports
that use the SROI [37] as a measurement method.
Although this number of reports may be insufficient
to obtain optimal results via a text mining process, we
consider that they are sufficient to state and validate
our hypotheses. In the future, we are contemplating
5

http://www.socialvalueuk.org/

March 2021

Figure 1. SIAMES frame.

the incorporation of the reports that Social Value U.K.
continues collecting on its website and through its
project Global Value Exchange (GVE)6 as they will
contribute greater wealth to our corpus.
The difficulty in finding enough number of reports
to be able to apply information extraction methods
and the scarcity of freely available quality reports
explain why the corpus is made up documents very
heterogeneous, both in their structure and writing
and in the topics they deal with. This diversity of
reports, although it is a problem when it comes to
extracting information, will serve to cover a greater
number of cases by nourishing our system with a
broad base.
6

http://www.globalvaluexchange.org/

http://www.socialvalueuk.org/ http://www.globalvaluexchange.org/

IEEE Technology and Society Magazine - March 2021

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - March 2021