Sandro Pezzelle

The LAMBADA dataset: Word prediction requiring a broad discourse context

We introduce LAMBADA, a dataset to evaluate the capabilities of computational models for text understanding by means of a word prediction task. LAMBADA is a collection of narrative passages sharing the characteristic that human subjects are able to guess their last word if they are exposed to the whole passage, but not if they only see the last sentence preceding the target word. To succeed on LAMBADA, computational models cannot simply rely on local context, but must be able to keep track of information in the broader discourse. We show that LAMBADA exemplifies a wide range of linguistic phenomena, and that none of several state-of-the-art language models reaches accuracy above 1% on this novel benchmark. We thus propose LAMBADA as a challenging test set, meant to encourage the development of new models capable of genuine understanding of broad context in natural language text.

“Look, some green circles!”: learning to quantify from images

In this paper, we investigate whether a neural network model can learn the meaning of natural language quantifiers (no, some and all) from their use in visual contexts. We show that memory networks perform well in this task, and that explicit counting is not necessary to the system’s performance, supporting psycholinguistic evidence on the acquisition of quantifiers.

Building a Bagpipe with a Bag and a Pipe: Exploring Conceptual Combination in Vision.

This preliminary study investigates whether, and to what extent, conceptual combination is conveyed by vision. Working with noun-noun compounds we show that, for some cases, the composed visual vector built with a simple additive model is effective in approximating the visual vector representing the complex concept.

Imparare a quantificare guardando

English. In this paper, we focus on linguistic questions over images which may be answered with a quantifier (e.g. How many dogs are black? Some/most/all of them, etc.). We show that in order to learn to quantify, a multimodal model has to obtain a genuine understanding of linguistic and visual inputs and of their interaction. We propose a model that extracts a fuzzy representation of the set of the queried objects (e.g. dogs) and of the queried property in relation to that set (e.g. black with respect to dogs), outputting the appropriate quantifier for that relation. Italiano. In questo lavoro studiamo le domande del tipo “Quanti cani sono neri?”, la cui risposta è un quantificatore (es. “alcuni”/”tutti”/”nessuno”). Mostriamo che al fine di imparare a quantificare, un modello multimodale deve ottenere una rappresentazione profonda della domanda linguistica, dell’immagine e della loro interazione. Proponiamo un modello che estrae una rappresentazione approssimativa dell’insieme degli oggetti e della proprietà sui quali verte la domanda.