The proposed methods often model entity mentions as nodes and relations as directed arcs with relation types as labels. Then they apply some linear models or deep models to solve it.

University of Michigan - Ann Arbor

There are three kinds of major joint models used on Relation Extraction:
1. Structure Prediction
2. Neural Models
3. Table Representation

Joint Models of Relation Extraction

Structured prediction or structured (output) learning is an umbrella term for supervised machine learning techniques that involves predicting structured objects, rather than scalar discrete or real values.

Structured Prediction

Techniques for Jointly Extracting Entities and Relations: A Survey

Structured Prediction for Relation Extraction

Predictions for both the tasks of entity and relation extraction are carried out using a single joint neural model, where at least some of the model parameters are shared across both tasks. Joint modeling is realized through such parameter sharing where training for any task updates the parameters involved in both the tasks.

Neural Models for Relation Extraction

A table is associated with each sentence where every table cell is labelled with an appropriate label so that the whole entities and relations structure in a sentence is represented in a single table.

Learn Before

Related