Entity Typing is an important task in text analysis. Assigning types (e.g., person, location, organization) to mentions of entities in documents enables effective structured analysis of unstructured text corpora. The extracted type information can be used in a wide range of ways (e.g., serving as primitives for information extraction and knowledge base (KB) completion, and assisting question answering). Traditional Entity Typing systems focus on a small set of coarse types (typically fewer than 10). Recent studies work on a much larger set of fine-grained types which form a tree-structured hierarchy (e.g., actor as a subtype of artist, and artist is a subtype of person).
Source: Label Noise Reduction in Entity Typing by Heterogeneous Partial-Label Embedding
Image Credit: Label Noise Reduction in Entity Typing by Heterogeneous Partial-Label Embedding
Neural language representation models such as BERT pre-trained on large-scale corpora can well capture rich semantic patterns from plain text, and be fine-tuned to consistently improve the performance of various NLP tasks.
Ranked #1 on
Entity Linking
on FIGER
ENTITY LINKING ENTITY TYPING KNOWLEDGE GRAPHS LINGUISTIC ACCEPTABILITY NATURAL LANGUAGE INFERENCE PARAPHRASE IDENTIFICATION SENTIMENT ANALYSIS
We present BPEmb, a collection of pre-trained subword unit embeddings in 275 languages, based on Byte-Pair Encoding (BPE).
In this paper, we describe TextEnt, a neural network model that learns distributed representations of entities and documents directly from a knowledge base (KB).
Ranked #1 on
Entity Typing
on Freebase FIGER
In this paper, we propose new pretrained contextualized representations of words and entities based on the bidirectional transformer.
Ranked #1 on
Question Answering
on TACRED
ENTITY TYPING LANGUAGE MODELLING NAMED ENTITY RECOGNITION QUESTION ANSWERING RELATION CLASSIFICATION
We study the problem of injecting knowledge into large pre-trained models like BERT and RoBERTa.
Ranked #1 on
Entity Typing
on Open Entity
DEPENDENCY PARSING ENTITY TYPING QUESTION ANSWERING RELATION CLASSIFICATION
Contextual word representations, typically trained on unstructured, unlabeled text, do not contain any explicit grounding to real world entities and are often unable to remember facts about those entities.
Ranked #6 on
Relation Extraction
on TACRED
(using extra training data)
ENTITY LINKING ENTITY TYPING LANGUAGE MODELLING RELATION EXTRACTION WORD SENSE DISAMBIGUATION
Extraction from raw text to a knowledge base of entities and fine-grained types is often cast as prediction into a flat set of entity and type labels, neglecting the rich hierarchies over types and entities contained in curated ontologies.
Our model not only utilizes entities and their latent types as features effectively but also is more interpretable by visualizing attention mechanisms applied to our model and results of LET.
Ranked #13 on
Relation Extraction
on SemEval-2010 Task 8
Existing entity typing systems usually exploit the type hierarchy provided by knowledge base (KB) schema to model label correlations and thus improve the overall performance.
Ranked #2 on
Entity Typing
on Ontonotes v5 (English)
Current systems of fine-grained entity typing use distant supervision in conjunction with existing knowledge bases to assign categories (type labels) to entity mentions.
ENTITY TYPING SEMANTIC SIMILARITY SEMANTIC TEXTUAL SIMILARITY