Entity-Knowledge Ontologies

We have created a large entity-knowledge ontology (using Virtuoso) which can incorporate basic facts from Wikipedia, WordNet and other freely-available knowledge sources. The ontology can be used to generate questions on any of its entities (e.g., people, places, events) or to discover facts linking two or more entities together.  A host of front-end tools have been developed that provide visualization capabilities, which allow for the manipulation of this very large set of data (editing, fact-finding) and the automatic generation of questions. The main area of application is providing content for a recently-launched quiz-based app which is extremely rich in variety and information content compared to its peers.

Automatic Question Generation (EDUIZ ®, Quizedia™, Flooxi)

We have applied ideas from human-computer interaction and CBA/CBT, together with text and data mining techniques, neural networks and genetic algorithms, towards the automated creation of multiple-choice questions. Our team has identified a series of methods of creating questions of any type according to the tutor’s needs, the user’s ability, the required scope etc.

Our team has made the first attempt (to our knowledge) to create a solid, expandable and automated e-learning CBA/CBT system that combines data from online libraries and encyclopedias with text-mining and artificial intelligence tools and algorithms.

Quiz Generation Engine

We have developed software for the real-time automatic generation of quizzes conforming to well-defined criteria of variety, difficulty, etc. to be played online by two opponents. For this project we use tree-like structures, advanced caching mechanisms and other well-defined data structures to pre-load questions based on certain properties. We have also implemented an ELO rating system, similarly to FIDE in chess.

We employ genetic algorithms to ensure the quizzes produced fit the requirements for both players (P2P) using output from collaborative filtering algorithms. This is a clustered, sharded operation. Our fitness function requires – among other things - that users will not see duplicate questions, which is achieved using enhanced dispersion metrics.

Server Security and Big Data Synchronization

Got Malware? Does your site require SSL? What about encrypted databases and passwordless login? Safemarket offers hassle-free professional malware clean up, Digital Certificate Management, Server Administration, custom database synchronizers and passwordless access tools. We do it all for you, so no need for the extra burden on your resources.

TRE Development

Time-release encryption (TRE) is the encryption of confidential data so that the resulting encrypted text cannot be deciphered by anyone, including the intended recipient, until a predetermined future time. TRE could be useful, and perhaps essential, in many real world situations with special relevance to e-governance applications, where the information transmitted in advance must remain secure for a specific time period.


Among the best-known examples are e-voting, where the delayed opening of votes is required; sealed-bid auctions where offers must remain sealed until the end of bidding; timed-release of electronic documents; e-contests, where participants must not have access to the question before the contest begins. Other applications include scheduled payments, timestamps and contract signing. Research on this topic has produced a wealth of theoretical developments and provable secure protocols, some the most efficient being Cathalo et al, 2005 and Chalkias et al, 2007. Moreover, an implementation infrastructure has been proposed in Chalkias et al, 2008 which will be the base for our construction. It is interesting that note that although TRE has been recently proposed for the applications mentioned above, it has not been implemented in practice. TRE has the potential to be a highly useful tool with significance for the operation of large-scale contests such as national examinations.

Verifiable Random Functions

“A 51-year old was arrested in January 2015 and charged with two counts of fraud after being suspected of tampering with the lottery computers that generate random numbers. He was working as the information security director at the Multi-State Lottery Association when he was captured on CCTV purchasing a Hot Lotto ticket.” (link)

We believe it is time to make events like this impossible by creating a “self-verifiable random beacon”, i.e., a random function which is tamper-proof and can be verified by anyone. Towards that end, we have developed SafeRandom, a patent pending mechanism for generating secure, fair, verifiable and transparent random numbers! For further questions regarding this project please contact us.