Same Language Translation
Mar 3rd
Dr. Peter Cahill, Prof. Julie Carson-Berndsen of the School of Computer Science and Informatics are active members of the SFI funded Centre for Next Generation Localisation CSET. Along Jinhua Du, Andy Way of School of Computing, Dublin City University they have developed a system for undestandability of text using Same Language Translation.
In their invention machine translation has been turned to the task of making a language (the same language for input and output) more understandable while retaining the original context. A key benefit offered by the implementation of this invention is that “naturalness” of machine translation is no longer limited by the training data used. The system chooses how to speak a piece if text based on the training data it has and the retention of meaning based on context.
For more information contact Peter Conlon. A marketing flyer can be downloaded here (pdf).
FORTRESS: Filtering Hidden Information in Network Data
Feb 3rd
Dr. Guenole Silvestre and Dr. Neil Hurley of the UCD School of Computer Science & Informatics have developed FORTRESS an efficient tool for Filtering Hidden Information in Network Data. FORTRESS (FORensic Tool for Resilient Steganographic Signatures) provides a reliable, low latency method for withholding unauthorised content from Internet users while being easy to use like an anti-virus filter.
Steganography is the ancient art of hiding information. Digital technology gives new ways to hide information in multimedia content such that it remains perceptually unchanged. An example is digital watermarking which balances the need to achieve statistical transparency while maximizing the amount of transmitted information.
FORTRESS can be deployed to detect hidden information amongst otherwise innocuous Internet traffic. This hidden information may comprise of illegal or offensive images such as pornography, terrorist messages or unauthorized data such as industrial espionage.
For more information contact Peter Conlon. A marketing flyer can be downloaded here (pdf).
Fast and Scalable Derivatives Pricing
Jan 27th
Dr. Conall O’Sullivan of The School of Business and his colleague Dr. Stephen O’Sullivan (now with the Dublin Institute of Technology) are recognised experts in the mathematically challenging world of Complex Derivatives Pricing (CDP). Over the last two years they have conducted research into alternative computational methods for CDP. These new methods achieve similar levels of pricing accuracy, significant improvements in computational speed but with more flexibility and generality of implementation. Their new approach allows implementations to be easily parallized leveraging multi-core processors, something that current methods can’t easily achieve.
This Fast and Scalable Derivatives Pricing technology is ready for productization. The teams next objective is to conduct a market feasibility study to determine the size and type of the potential market. This will involve the derivatives desks and research departments of the major financial companies who are the main institutions brokering complex derivatives deals with investment banks and hedge funds.
Interested? For more information contact Peter Conlon.
For Researchers – Delivering Software Projects
Jan 25th
When developing a prototype/technology demonstrator it’s important to arrive at the Minimum Viable Product (MVP). The MVP defines the least amount of development that you have to do to deliver a product that a customer will buy.
First off you have to gather requirements (owned by the customer and marketing team). These requirements drive a set product features (negotiated between marketing and the development team). Each of the features will in turn be defined as a set of specifications (owned by the development team but heavily referenced back to the requirements. Each feature will have a performance metric attached that will dictate when aand if a feature passes its acceptance tests). Do this up front and validate with your customer. Agree that the test performance metrics will satisfy their requirements.
Innovation Reading List
Jan 11th
I recently re-discovered a reading list for “Innovation Practitioners” that you might find useful.
Identifying Business Development Opportunities
Jan 7th
All good Universities will have researchers investigating many different and discrete areas of interest. Its easy to treat each invention or discovery that is made in the University as a point instance of a particular technology. However this leads to a lot of duplication of effort and individual technologies that are hard to market individually.
Outside of the major research initiatives such as Clarity and Clique, UCD has a breath and depth of expertise in a number of key areas of ICT. To help the process of building up a managed portfolio of commercialisation IP I have identified 3 broad areas:
- Wireless communications and networking – Physical infrastructure, low power signalling, protocols, management/routing etc.
- Information Control – Data forensics, information hiding/steganography, network analysis, encryption etc.
- Teaching and Learning – in particular spoken and written language
Other possible areas could include energy/utility management and medical imaging.
Watch this space for future posts expanding in the various business development opportunities I identify.
FastTRACT: visualizing neural pathways in the brain.
Dec 11th
Dr. Kathleen Curran, Eoin Murphy, Stephen Meredith and Niall Colgan of the School of Medicine & Medical Science have developed an easy to use visualization platform for Diffusion Tensor Imaging data.
Diffusion Tensor Imaging can provide a unique insight into the neuronal architecture of the human brain in-vivo. A significant barrier to the clinical utilisation of tractography is that fibre pathways generated differ from true anatomical connectivity.
FastTRACT is standalone software tool specifically developed in collaboration with consultant radiologists, neurologists and neurosurgeons to address the variability of results and reliability of the reconstructed fibre representations.
For more information contact Peter Conlon. A pre-marketing flyer can be downloaded here (pdf).
Following the P2P file sharing trail
Dec 8th
Two PhD students at the UCD Centre for Cybercrime Investigation are currently investigating novel ways to discover and model the behaviour of P2P file sharing networks. There are currently working on a Proof of Concept demonstrator which at this stage is capable of tracking how the file sharing network is created, where the key nodes are, who is participating and whats its life time etc. After the fact investigation of P2P networks is what is currently being concentrated on but its easy to see applications in enforcement of copyright laws and network P2P monitoring for QOS maintenance.
Contact Peter Conlon for more information and watch out for future posts
Using micro-blogging to Recommend Real-Time Topical News
Nov 10th
Owen Phelan, Dr. Kevin McCarthy & Prof. Barry Smyth, CLARITY: Centre for Sensor Web Technologies, School of Computer Science and Informatics have developed a news recommendation system based on micro-blogging and RSS feeds.
With the growing popularity of social networking, RSS feeds are being accessed less and less. The proposed technology marries the depth of news from RSS feeds with the breath of interests from Social Networking. News items are recommended as the user interacts others via micro-blogging. Targeted delivery of news allows for opportunistic deliver of advertisements tuned to what interests the user (as opposed to what they are searching for).
For more information contact Peter Conlon. A marketing flyer can be downloaded here (pdf).
Digital pre-distortion for RF power amplifier for wireless communication
Nov 9th
Dr. Anding Zhu and Lei Guan of UCD School of Electrical, Electronic and Mechanical Engineering have developed a digital predistortor (DPD) for RF power amplification that:
-
Offers a significant reduction in DPD implementation complexity
-
Compensates for PA non-linearity and memory effects
-
Is a DPD design that can be implemented on a single FPGA
-
Control parameters can changed in memory without having to modify the DPD implementation
For more information, including papers, contact Peter Conlon. Marketing flyer can be downloaded here (pdf).