Invited Talks
|
Learning With Explanations
Tim Rocktäschel Despite the success of deep learning models in a wide range of applications, these methods suffer from low sample efficiency and opaqueness. Low sample efficiency limits the application of deep learning to domains for which abundant training data exists whereas opaqueness prevents us from understanding how a model derived a particular output, let alone how to correct systematic errors, how to remove bias, or how to incorporate common sense and domain knowledge. To address these issues for knowledge base completion, we developed end-to-end differentiable provers which (i) learn neural representations of symbols in a knowledge base, (ii) make use of similarities between learned symbol representations to prove queries to the knowledge base, (iii) induce logical rules, and (iv) use provided and induced rules for multi-hop reasoning. I will present our recent efforts in applying differentiable provers to statements in natural language texts and large-scale knowledge bases. Furthermore, I will introduce two datasets for advancing the development of models capable of incorporating natural language explanations: eSNLI, crowdsourced explanations for over half a million sentence pairs in the Stanford Natural Language Inference corpus, and ShARC, a conversational question answering dataset with natural language rules. |
|
Argumentation Mining and Generation Supporting the Verification of Content
Marie-Francine Moens Argumentation mining and generation are intelligent tasks mastered by humans. They both require understanding of human language in context. In this talk we give an overview of state-of-the-art argumentation mining and generation techniques, and we show their potential in the verification of content. We elaborate on how to extract facts or claims from text and their supporting and non-supporting arguments. In order to verify content, the resulting content representations are compared with those of facts and arguments extracted from other texts and possibly extracted from other sources, or with those that are available in knowledge repositories. Current research starts to explore the automated generation of arguments given a certain claim and its context, which might lead to ways that facilitate the verification of claims. |
|
Building a broad knowledge graph for products
Luna Dong Knowledge graphs have been used to support a wide range of applications and enhance search results for multiple major search engines, such as Google and Bing. At Amazon we are building a Product Graph, an authoritative knowledge graph for all products in the world. The thousands of product verticals we need to model, the vast number of data sources we need to extract knowledge from, the huge volume of new products we need to handle every day, and the various applications in Search, Discovery, Personalization, Voice, that we wish to support, all present big challenges in constructing such a graph. In this talk we describe our efforts in building a broad product graph, a graph that starts shallow with core entities and relationships, and allows easily adding verticals and relationships in a pay-as-you-go fashion. We describe our efforts on knowledge extraction, linkage, and cleaning to significantly improve the coverage and quality of product knowledge. We also present our progress towards our moon-shot goals including harvesting knowledge from the web, hands-off-the-wheel knowledge integration and cleaning, human-in-the-loop knowledge learning, and graph mining and graph-enhanced search. |
|
Call for Help: Putting Computation in Computational Fact Checking
Delip Rao Fact Checking, a discipline at least as old as journalism, is undergoing a massive overhaul to meet the needs of rapidly emergent news, digital consumption and social media, and the rise of misinformation and disinformation attacks. In this talk, I will review some of the attack vectors, major challenges faced by fact checkers and journalists and what we can offer as a community to help maintain the integrity of our news sources. I will also review some current efforts, future directions, and open problems related to computational fact checking. |
