We create Innovation

Whether you are a company or an individual, we will help you create a digital dimension to your work. As part of our continuous improvement, we also focus on modern IT technologies.

Meet customer needs

Innovation is essential today

Stand out from competitors

Process of creating value

We can create new value for customers

It is important to be clear about the difference between invention and innovation. Invention is a new idea. Innovation is the commercial application and successful exploitation of the idea.

Whether you are a company or an individual, we will help you create a digital dimension to your work. As part of our continuous improvement, we also focus on modern IT technologies. Our team’s current research involving Artificial IntelligenceCybersecurity and Internet of Things materialised in the prestigious prize PatriotFest 2018. We invite you to read the description check the winning project: 

http://www.cybersecurityanalyzer.com/

Artificial Intelligence

Artificial Intelligence is one of our many passions. We are currently working in research projects including Machine Learning and Big Data solutions. Tell us about your idea.

Security Services

Cybersecurity is one of the hottest topics in IT nowadays. We have experience both in penetration testing and security research. We are ready to find solutions to your software security concerns.

Also, we have research Experience, our main fields of interest are cybersecurity, machine learning and natural language processing.

Artificial Intelligence

Cybersecurity Analyzer has the ability to understand documents related to cybersecurity.

Machine Learning

Similar to a humai being, our robot becomes smarter as it is trained.

Award

Cybersecurity Analyzer was awarded the prestigious prize PatriotFest 2018.

Research

The solution is the result of 3 years of research conducted in The Bucharest University of Economic Studies.

Cybersecurity Analyzer is an innovative software solution that analyzes and understands documents about cybersecurity.

Built on an ontology which contains over 10.000 elements, the solution is able to identify relevant aspects related to cybersecurity. Cybersecurity Analyzer is an Artificial Intelligence solution. Similar to a human being, it has the ability to understand text, identifying classes of concepts, entities and the relations between them.

The software solution is based on Machine Learning algorithms, becoming more powerful as it is trained. Using the obtained results, feedback is extracted and the model is automatically improved. Cybersecurity Analyzer has already been trained with documents which contains over 300,000 words.

Cybersecurity Analyzer can be adapted to automatically detect and interpret hacker discussions. The solution can also be used as a tool for information and learning for both cybersecurity experts and anyone interested in using the computer or other intelligent devices safely.

OUR WORK

Hotel Review Insights

Hotel review is a Machine Learning model designed to perform named entity recognition (NER), relation extraction (RE) and sentiment analysis (SA) for online reviews regarding hotels and resorts written in English.

The main purpose of our work consists in developing a NLP model which automatically analyses reviews related to hotels and resorts. In this regard, we studied related papers and identified the main methods, techniques and instruments used. We started by developing a domain ontology specific to the hotel industry field.

Regarding the ML approach, supervised learning was preferred, as in recent years studies show that it provides the best performance for NLP.
Knowledge Studio was used to define the model’s structure, to configure it according to the domain particularities and to train the model, while Natural Language Understanding was used to deploy the model and served as a gateway to the analysis results provided by our model.

Supervised ML models require the provision of correct result sets on which the machine builds the ground truth. In the training process, we annotated the documents for two tasks NER and RE. For NER, the annotation process consisted in identifying and labeling the relevant tokens and assigning each one to the proper class. In order to automate the process, the rule-based model was used as it automatically identified the tokens defined in the dictionary. However, much attention from the annotator was required, as many of the tokens weren’t correctly labeled by the dictionary.

For training, we downloaded reviews available online. We used documents totalizing approximately 100,000 words. In order to speed up the process, we developed custom-made scraping solutions to automatically download and structure the relevant data.

The user can also get a more detailed look on one class’ sentiment score by clicking on the corresponding button and consulting the table of identified entities and associated sentiment score.

Let's Work