Seminari CoLing-Lab, 15 marzo


    Mercoledì 15 marzo presso la Sala Riunioni di Pal. Venera (Via S. Maria 36, secondo piano), si terranno i seguenti seminari:
    • ore 10:00-11:00 –  Prof. Chris Biemann (Università di Amburgo)
    Adaptive Natural Language Processing with Graph-based Distributional Semantics
    Abstract
    In this talk, I will motivate an adaptive approach to natural language processing, where NLP components get smarter through usage over time, following a ‘cognitive computing’ approach to natural language processing. With the help of recent research prototypes, three stages of data-driven adaptation will be illustrated: feature/resource induction, induction of processing components and continuous data-driven learning.
    Many of these approaches have been implemented with a scalable graph-based solution to distributional semantics, which belongs to the family of ’count-based’ DSMs, keeps its representation sparse and explicit, and thus fully interpretable. To put this into perspective, I will highlight some important differences between sparse graph-based and dense vector approaches to DSMs: while dense vector-based models are computationally easier to handle and provide a nice uniform representation, they lack interpretability, provenance and robustness. On the other hand, graph-based sparse models have a more straightforward interpretation, handle sense distinctions more naturally and can straightforwardly be linked to knowledge bases, while lacking the ability to compare arbitrary lexical units and a compositionality operation. Since both representations have their merits, I opt for exploring their combination in the outlook.
    • ore 11:30-12:30 – Prof. Alessandro Lenci (Università di Pisa)
    Distributional Semantics and Sentence Comprehension
    Abstract
    In this talk, I argue that the comprehension of a sentence is an incremental process driven by the goal of constructing a coherent representation of the event the speaker intends to communicate. I will introduce a distributional model to build semantic representations inspired by recent psycholinguistic research on sentence comprehension. The model also associates with each distributional semantic representation a composition cost, to model the cognitive effort necessary to build it. The composition cost depends on the internal coherence of the event representation being constructed and on the activation degree of such event by linguistic constructions. The model is tested on some psycholinguistic datasets for the study of sentence comprehension. In particular, I focus on the case of logical metonymy  (e.g. The student began the book) showing that the model can account for the extra processing cost of coercion and for the inferential process leading to the recovery of the implicit event.
     
    Chris Biemann obtained his doctorate in Computer Science/Natural Language Processing in 2007 from the University of Leipzig, before joining the San-Francisco-based semantic search startup Powerset, which was acquired by Microsoft to form the Bing.com search engine. In 2011, he got appointed as assistant professor for language technology in the computer science department at TU Darmstadt; since October 2016, Chris Biemann is professor for language technology at the University of Hamburg. His current research is focused on adaptive natural language processing in the cognitive computing paradigm, web-scale statistical semantic methods, machine learning from crowdsourcing signals and on applications in the humanities and social sciences.
    Pubblicato il 14 marzo 2017