WHAT WE DO

Intelligenza artificiale

QBT develops algorithms and software in the fintech area, backed by a proof experience in the direct provision of services.
The mentioned experience has increased our know-how, consolidating our ability to develop tailored and customized management softwares in order to provide the best technological solution according to the specific customers’ needs.
Moreover, through a deep research and development activity we explored new areas of business turning into a point of reference in the field of Artificial Intelligence, and, in particular, in the Natural Language Processing and machine learning.
We develop all of our research activities thanks to a close cooperation with universities and research institutes, which represent the natural competence network at QBT.

Artificial Intelligence:
Natural Language Processing e Machine Learning

«The Artificial Intelligence (known with the English acronym AI) is a discipline belonging to the computer science which studies theories, methodologies and techniques allowing the design of hardware systems and software programs to supply the electronic computer with services that, to an ordinary observer, would seem to be exclusively human».

(Marco Somalvico)

More

Specific definitions can be given by focusing on internal reasoning processes or on the external behavior of the intelligent system and using the similarity with human behavior or with an ideal behavior, called rational, as a measure of effectiveness.

The AI can simulate human intelligence in different ways, according to the processes activated:

  • acting humanly – the result of the operation performed by the intelligent system cannot be distinguished by the one belonging to the humankind;
  • thinking humanly – the process leading the intelligent system to solve a problem is very similar to the human one. This approach is linked to cognitive science;
  • thinking rationally – the process leading the intelligent system to solve a problem is a formal procedure referring to logic;
  • acting rationally - the process leading the intelligent system to solve a problem is the one that allows getting the best-expected result given the information available.

(source: Wikipedia)


Agent MOrSe 

Semantic research or Natural Language Processing:

  • does not rely on the simple search of Keywords;
  • solves the problems linked to morphology (singular/plural, infinitive verbs, etc.);
  • understands the meaning of the “context” thought, through the disambiguation of texts;
  • provides results that refer to the context even if the Keywords are not specifically present in the content.

An example is shown below:

The search “Penguins” will provide all the information sources reporting contents related to the funny “grey animals of the Southern Pole”, without such contents having to include strictly the word “Penguins”.


MOrSe 
(Semantic Search Engine) is a QBT platform, based on proprietary semantic technology, which supports customers in managing the information available in the most effective way, to get fundamental and strategic notions.
The platform can manage large data flows (text documents, multimedia, audio streams, web pages and social network) in 27 different languages.
The strength of the MOrSe platform are the specific thesauri defined for each specific application sector to conceptualize, classify and search for topics in the best way. This allows going beyond the limits of traditional technologies, which are based on keyword.
Designed and realized by experts of the sector together with CNR in Rome, MOrSe is adopted by qualified customers in Italy and Switzerland and it is used in different and multiple environments.

Twitter Opinion Dynamic

Twitter Opinion Dynamic

Twitter Opinion Dynamic is a tweet analysis algorithm pointing out the opinion of the users of the social network Twitter on specific topics, through the so-called sentiment analysis.
The advantages of this approach rely on high speed and responsiveness (the source is the web) which allow studying and monitoring the effects of the studied events.

The potential of this kind of analysis is big and as follows:

  • it can be done quickly and at any time;
  • the system is trained only once in extremely short times and directly by the user, even if there is no statistics knowledge, but familiarity with the topic under monitoring;
  • the analysis is not on a sample but on the opinion emerging from the analysis of the sources, which can be mediated (newspapers) or non-mediated (blog, forum, etc.);
  • it is objective because it does not ask questions but remains “listening”.

In this approach, it is fundamental to study the internet user who posts contents on the web: compared to the survey, in fact, the study of the internet is affected by an intrinsic lack of coverage in the sampling, as, for instance, not all social categories have access to the web with adequate numbers to generate representativeness.
Our Sentiment Analysis thus categorizes the web users and then analyzes their opinion, in order to generate Opinion Dynamic analysis, ranging from politics to understanding social phenomena or market trends.
The model allows the external measurement of reputational risk (of qualitative kind) among different stakeholders, providing management with an effective tool to monitor the risk in a dynamic way.

The tool is available at http://www.tody.it