Abstract

Abstract

Abstract
Modern search is heavily powered by knowledge bases, but users still query using keywords or natural language. As search becomes increasingly dependent on the integration of text and knowledge, novel approaches for a unified representation of combined data present the opportunity to unlock new ranking strategies. We have previously proposed the graph-of-entity as a purely graph-based representation and retrieval model, however this model would scale poorly. We tackle the scalability issue by adapting the model so that it can be represented as a hypergraph. This enables a significant reduction of the number of (hyper)edges, in regard to the number of nodes, while nearly capturing the same amount of information. Moreover, such a higher-order data structure, presents the ability to capture richer types of relations, including nary connections such as synonymy, or subsumption. We present the hypergraph-of-entity as the next step in the graph-of-entity model, where we explore a ranking approach based on biased random walks. We evaluate the approaches using a subset of the INEX 2009 Wikipedia Collection. While performance is still below the state of the art, we were, in part, able to achieve a MAP score similar to TF-IDF and greatly improve indexing efficiency over the graph-of-entity.

Abstract

Abstract
Managing large volumes of digital documents along with the information they contain, or are associated with, can be challenging. As systems become more intelligent, it increasingly makes sense to power retrieval through all available data, where every lead makes it easier to reach relevant documents or entities. Modern search is heavily powered by structured knowledge, but users still query using keywords or, at the very best, telegraphic natural language. As search becomes increasingly dependent on the integration of text and knowledge, novel approaches for a unified representation of combined data present the opportunity to unlock new ranking strategies. We tackle entity-oriented search using graph-based approaches for representation and retrieval. In particular, we propose the graph-of-entity, a novel approach for indexing combined data, where terms, entities and their relations are jointly represented. We compare the graph-of-entity with the graph-of-word, a text-only model, verifying that, overall, it does not yet achieve a better performance, despite obtaining a higher precision. Our assessment was based on a small subset of the INEX 2009 Wikipedia Collection, created from a sample of 10 topics and respectively judged documents. The offline evaluation we do here is complementary to its counterpart from TREC 2017 OpenSearch track, where, during our participation, we had assessed graph-of-entity in an online setting, through team-draft interleaving. © José Devezas, Carla Lopes, and Sérgio Nunes.

Abstract
As entity-oriented search takes the lead in modern search, the need for increasingly flexible tools, capable of motivating innovation in information retrieval research, also becomes more evident. Army ANT is an open source framework that takes a step forward in generalizing information retrieval research, so that modern approaches can be easily integrated in a shared evaluation environment. We present an overview on the system architecture of Army ANT, which has four main abstractions: (i) readers, to iterate over text collections, potentially containing associated entities and triples; (ii) engines, that implement indexing and searching approaches, supporting different retrieval tasks and ranking functions; (iii) databases, to store additional document metadata; and (iv) evaluators, to assess retrieval performance for specific tasks and test collections. We also introduce the command line interface and the web interface, presenting a learn mode as a way to explore, analyze and understand representation and retrieval models, through tracing, score component visualization and documentation. © Springer Nature Switzerland AG 2020.