Neural dialogue models: growing up confident and independent

2:00 PM

Neural approaches to dialogue modelling have produced state of the art results in each dialogue subtask, ranging from dialogue state tracking to user modelling. One of the biggest advantages of neural architectures for dialogue is their potential for end-to-end dialogue modelling. However, these, in many ways powerful, models have their Achilles’ heels. Their weaknesses are particularly apparent in three learning aspects, namely independence, confidence and interaction. In this talk, I will show how each of these aspects can be addressed in neural approaches to dialogue state tracking, dialogue policy estimation and user modelling.

What we learn from Google n-grams about the formation of new morphological paradigms? The case of Italian N + N verbal-nexus composites

2:00 PM

N + N composites are a recent structural innovation in Romance languages, the diachronic development of which has not yet been mapped in more detail. The aim of the presented empirical diachronic research is to specify the origin and mechanism of formation of this word-forming formula with a theoretical anchor in the paradigm of structural and relational morphology. For a detailed mapping of the diachronic development of Italian verbal-nexus composites, we tried to use data from Google n-grams (version 2020) and Google books. Analyzes of a sample of more than 1,000 composites (types) so far show that the origin of this formula is probably much older and can be traced in the official language since at least the middle of the 19th century. In the lecture and subsequent discussion, we would like to focus primarily on the technical and methodological aspects of research.

Multilingual Coreference Resolution

2:00 PM

We will describe an end-to-end coreference resolution system and experiments on recently created multilingual corpus CorefUD. In addition to monolingual experiments, we combine the training data in multilingual experiments and train two joined models – for Slavic languages and all the languages together. We rely on an end-to-end deep learning model adapted for the CorefUD corpus. We discuss the difficulties we faced, mainly regarding the differences between corpora in different languages. Next, we focus on the problem of predicting singleton relations. Finally, we deal with the benefits of harmonized annotations. We will show that using joined models helps significantly for the languages with smaller training data.

Coding da Vinci – Kick-Off weekend

Coding da Vinci is the first so-called hackathon, which focuses on the use of open data in the field of culture and cultural heritage. It is a platform for those who are interested in the field of open data and want to participate in the creation of innovative projects through applications, websites, the use of VR / AR, interactive installations, data visualizations.

Versology and Authorship Determination - Petr Plecháč (Institute of Czech Literature AS CR, v.v.i.)

2:00 PM

The lecture will focus on the extent to which the characteristics of verse and rhymes can be used to determine authorship. The lecturer will then present specific applications of versological attribution to controversial texts attributed to W. Shakespeare and G. Batěnkov.

Opening exhibition of Vasulka Live Archive / Interfaces

5:00 PM

Presentation of the Vasulka Live Archive project, which summarizes several years of cooperation between a team of art scientists and programmers to showcase the work of Stein and Woody Vašulková using artificial intelligence and interactive interfaces. The opening of the exhibition will be followed by an audiovisual performance by the British improvisation duo Yeah You, made up of a father and daughter - Mykl Jaxn and Elvin Brandhi; in Vašulka Kitchen Brno he will present his sample from a new board called KRUTCH.