Bootleg: Chasing the Tail with Self-Supervised Named Entity Disambiguation

Abstract

A challenge for named entity disambiguation (NED), the task of mapping textual mentions to entities in a knowledge base, is how to disambiguate entities that appear rarely in the training data, termed tail entities. Humans use subtle reasoning patterns based on knowledge of entity facts, relations, and types to disambiguate unfamiliar entities. Inspired by these patterns, we introduce Bootleg, a self-supervised NED system that is explicitly grounded in reasoning patterns for disambiguation. We define core reasoning patterns for disambiguation, create a learning procedure to encourage the self-supervised model to learn the patterns, and show how to use weak supervision to enhance the signals in the training data. Encoding the reasoning patterns in a simple Transformer architecture, Bootleg meets or exceeds state-of-the-art on three NED benchmarks. We further show that the learned representations from Bootleg successfully transfer to other non-disambiguation tasks that require entity-based knowledge: we set a new state-ofthe-art in the popular TACRED relation extraction task by 1.0 F1 points and demonstrate up to 8% performance lift in highly optimized production search and assistant tasks at a major technology company.

Metadata Shaping: A Simple Approach for Knowledge-Enhanced Language Models

“…This work explores how to effectively provide these properties to popular transformer models. Tail entities are those seen < 10 times during training and head entities are seen ≥ 10 times, consistent with Orr et al (2020); Goel et al (2021).…”

“…We collect entity metadata from Wikidata for our evaluations, a compelling choice as several works successfully improve tail performance in industrial workloads using the knowledge base (e.g., Orr et al (2020)) We use the state-of-the-art pretrained entity-linking model from Orr et al (2020) to link the text in each task to an October 2020 dump of Wikidata. We use Wikidata and the first sentence of an entity's Wikipedia page to obtain descriptions.…”

“…Recent language models (LMs) such as BERT and its successors are remarkable at memorizing knowledge seen frequently during training, however performance degrades over the long tail of rare facts. Given the importance of factual knowledge for tasks such as questionanswering, search, and personal assistants (Bernstein et al, 2012;Poerner et al, 2020;Orr et al, 2020), there has been significant interest in injecting these base LMs with factual knowledge about entities (Zhang et al, 2019;Peters et al, 2019, inter alia.). In this work, we work we propose a simple and effective approach for enhancing LMs with knowledge, called metadata shaping.…”

Ember: No-Code Context Enrichment via Similarity-Based Keyless Joins

“…We now describe a subset of these domains, which we evaluate in Section 8. Additional applicable domains include entity linkage or disambiguation [46,55], nearest neighbor machine translation [28], and nearest neighbor language modeling [29].…”

Evaluating Entity Disambiguation and the Role of Popularity in Retrieval-Based NLP

“…We use AmbER sets to conduct a systematic study of various retrieval systems that operate under different principles, such as token overlap and dense embedding similarity. Retrievers perform very differently on AmbER sets in terms of absolute retrieval numbers, with Bootleg (Orr et al, 2020), an entity-linking-based retriever, performing best. Despite these differences, all retrievers exhibit a large degree of popularity bias, underperforming on inputs concerning tail entities.…”