The annual social event for Informatics staff, students and their families took place on Friday 27 May.
The programme included a bake sale, poster session, quiz, coding workshop for children and a dinner and social, followed by a ceilidh.
Professor Mirella Lapata, Dr James Cheney and Professor Barbara Webb will each give a talk.
1pm - 2pm: Bake sale.
2pm - 3:30pm: Talks aimed at a School-wide audience (abstracts/speaker information below)
3.30pm - 4.30pm: Poster session and coffee.
4.30pm - 5.30pm: Informatics quiz.
5pm - 6pm: Coding workshop for children.
6.15pm - 11pm: Dinner and social followed by ceilidh at 7pm.
Recent years have witnessed the development of a wide range of computational methods and tools that process and generate natural language text. Many of these have become familiar to mainstream computer users such as tools that retrieve answers to questions, perform sentiment analysis, and translate between languages. The accessibility of the web could be further enhanced with applications that not only translate between different languages (e.g., from English to French) but also within the same language, between different modalities, or different data formats. The web is rife with non-linguistic data (e.g., databases, graphs, images, mathematical formulas, source code) that cannot be indexed or searched since most retrieval tools operate over textual data.
In this talk I will argue that in order to render electronic data more accessible to individuals and computers alike, new types of translation models need to be developed. I will sketch how recent advances in deep learning can be extended in order to induce general representations for different modalities and learn how these should be rendered in natural language.
Although great strides have been made in the last few years on software verification, it isn't yet possible to create production software for scientific data analysis that is correct and provably free of bugs. In some cases, it isn't even possible to specify what correctness means, for example, when the validity of the underlying scientific principles or models is subject to debate. Moreover, the most popular languages used for data analysis (to phrase it politely) don't meet the highest standards for language design. The inevitable resulting errors have already led to retractions of scientific articles and damage to careers. This is a problem for any of us who believe the word "science" in "data science" or "computer science" is worth taking seriously.
In this talk I'll elaborate on these problems and summarize some of my research on provenance, a small part of this problem: understanding the needs for accountability and validation of scientific computations and developing or analyzing proposed solutions.
The invertebrate brain is an evolutionary success story - it has capabilities that support robust adaptive behaviour (unmatched by any existing robots) in a huge variety of of tasks and environments. It is also of a size and scale (relative to the human brain) that we can potentially model and understand. In this short talk I will focus on recent work where we have been able to demonstrate a surprisingly direct connection between the precise neural wiring and the computational function of a specific area of the insect brain, potentially explaining their sophisticated navigational capacity for path integration.