Students' presentations

Adaptive Dialogue Models for Conversational Agents

Vevake Balaraman (Fondazione Bruno Kessler)

4 MAY 2018 - 11.45 - 12.45 Students' Presentations

Conversational agents exploit natural language technologies to engage users in text-based information seeking and task-oriented dialogs towards a specific domain. While there has been advances in the field to alleviate the need for large in-domain data, bootstrapping a competitive model for a new domain with no or very little data data is still a challenge. The new research direction of “zero-shot learning” addresses this problem of giving machines the ability to learn without training data for a domain. This presentation will focus on the need for such models and the recent advancements in the field of artificial intelligence towards achieving that.

Research on creativity: hapax, multiword expressions, constructions, Sketchengine

Fabio Loiodice (University of Bergamo-Pavia)

3 MAY 2018 - 16.45-17.45 Students' Presentations

Taking the notion of construction as a departure point we will try to define creativity as an analogic change of a given pattern. Afterwards we will use the sketchengine tools to find different types of datas: on one hand words, on the other hand syntagms, so that we will be able to provide a theoretical model of creativity that fits with the whole synyax-lexicon continuum.


Investigating morphological change through diachronic corpora

Silvia Micheli (University of Bergamo-Pavia)

3 MAY 2018 - 16.45-17.45 Students' Presentations

This presentation aims at showing how the evolution of a morphological construction can be investigated through a diachronic corpus of Italian. The study deals with a morphological construction made up of the string mal-, from the adverb male ‘badly’, plus a variable base (e.g. malsano, mal+sane ‘insane’; malconveniente, mal+convenient ‘inconvenient’; malpagato, mal+paid ‘low-paid’, malinterpretare, mal+interpret ‘to misinterpret’). Within these complex words, which are traditionally classified as compounds by dictionaries and grammars, mal- functions as a negative or evaluative prefix depending on the base meaning. The evolution of the mal-construction will be investigate on the basis of data extracted from a balanced diachronic corpus of Italian, i.e. the CoDIt_com (Corpus diacronico dell’italiano per la composizione). Its architecture as well as its limits and opportunities will be illustrated in respect of the goal of this research. The functional and distributional properties of the mal-construction will be discussed within the framework of Construction Morphology (Booij 2010).

A UD Literary Treebank for German. An Overview

Alessio Salomoni (University of Bergamo-Pavia)

4 MAY 2018 - 11.45 - 12.45 Students' Presentations

In this talk, I will outline the development of a historical dependency treebank for German. First, I will focus on the motivation behind this research project as well as on the choice of data. Then, I will depict the general methodology and the current status of the resource. A suggetion for linguistic analysis will be sketched in the end.