Abstract:The development of efficient methods for mining information from heterogeneous information networks (HINs) has become essential for improving the accuracy of collective entity linking in the absence of third-party knowledge bases. Currently, there remain three major challenges in the latest research: (1) The objective function of collective entity linking does not fully embrace the concept of "collective linking". (2) The objective function employs the mean value rather than the maximum value of the entity relatedness as a link parameter, while discounting the importance of the strong logical associations between the text and language for meaning recognition. (3) The objective function utilizes only one type of 2-hop path to contribute to entity relatedness, thereby disregarding other types of 2-hop path that exist in actual HINs. To address the aforementioned issues, this paper proposes a strong-relatedness-sequence-based fine-grained collective entity linking method (SRSCL). The SRSCL is capable of capturing the contextual information of the entity in the HINs, thereby providing improved accuracy in entity linking. Specifically, SRSCL constructs a knowledge representation learning model and proposes an overall semantic similarity model for entity mentions and candidate entities to solve the objective function and thereby reflect the idea of "collective linking". Additionally, a strong-relatedness-sequence-based overall relatedness measurement model is proposed for candidate entities to emphasize the strong logical associations between them. Furthermore, SRSCL defines three types of 2-hop path and evaluates the importance of each path to accurately measure the relatedness of entities. Finally, the experimental results demonstrate that the proposed SRSCL is more effective in capturing the overall relatedness of entities than the latest model. Particularly, when the number of entity mentions contained in one sliding window is greater than 6, the proposed SRSCL improves the precision, recall and F1 score by more than 10% compared with the latest model.

SHINE+: A General Framework for Domain-Specific Entity Linking with Heterogeneous Information Networks

A Probabilistic Model for Linking Named Entities in Web Text with Heterogeneous Information Networks

Boosting Collective Entity Linking via Type-Guided Semantic Embedding.

Cross-Lingual Entity Matching for Heterogeneous Online Wikis.

Toward Tweet Entity Linking with Heterogeneous Information Networks

SHINE: Signed Heterogeneous Information Network Embedding for Sentiment Link Prediction

SRSCL: A strong-relatedness-sequence-based fine-grained collective entity linking method for heterogeneous information networks

Xlink: An Unsupervised Bilingual Entity Linking System

Improving Entity Linking with Graph Networks

Source-Aware Embedding Training on Heterogeneous Information Networks

Entity Linking on Graph Data

Integrating Manifold Knowledge for Global Entity Linking with Heterogeneous Graphs

Entity Linking with a Knowledge Base: Issues, Techniques, and Solutions

LINDEN: linking named entities with knowledge base via semantic knowledge.

A collective entity linking algorithm with parallel computing on large-scale knowledge base

Class-Dynamic and Hierarchy-Constrained Network for Entity Linking

Learning a Probabilistic Semantic Model from Heterogeneous Social Networks for Relationship Identification

SHINE: SubHypergraph Inductive Neural nEtwork

A Lazy Learning Model for Entity Linking Using Query-Specific Information.

Entity Linking from Microblogs to Knowledge Base Using ListNet Algorithm.

Entity Linking Leveraging Automatically Generated Annotation.