Important
All Microsoft Academic Services have been officially retired as of December 31st, 2021. We are currently retaining original documentation as-is for educational use, however all information about signing up for services is no longer valid, and support and service (API) links will not function.
Network Similarity Package
The Microsoft Academic Network Similarity Package provides supplementary processing functionality for use with the Microsoft Academic Graph (MAG). For a detail description of the technology, please see Multi-Sense Network Representation Learning in Microsoft Academic Graph.
This package includes Python classes for Azure Databricks and U-SQL functions for Azure Data Lake Analytics. It also includes network embedding resources for sevaral academic entities similarity senses.
The functions/classes perform the following tasks.
- Similarity comparison between 2 entities using pre-trained network embeddings on the MAG corpus, and
- Compute top similar entities based on the pre-trained network embeddings.
Prerequisites
Before running these examples, you need to complete the following setups:
- Set up provisioning of Microsoft Academic Graph to an Azure blob storage account. See Get Microsoft Academic Graph on Azure storage.
Available senses
The following senses are of entity embeddings that are currently available.
| Entity Type | Sense | Description |
|---|---|---|
| affiliation | cofos | Two affiliations are similar if they publish papers with similar fields of study. |
| affiliation | copaper | Two affiliations are similar if they are closed connected with each other in the weighted affiliation collaboration graph. |
| affiliation | covenue | Two affiliations are similar if they publish in similar venues (journals and conferences). |
| affiliation | metapath | Two affiliations are similar if they co-occur with common affiliations, venues, and fields of study. |
| author | copaper | Two authors are similar if they are closed connected with each other in the weighted author collaboration graph. |
| conference | coauthor | Two conferences are similar if they publish papers with common authors. |
| conference | cofos | Two conferences are similar if they publish papers with similar fields of study. |
| conference | metapath | Two conferences are similar if they co-occur with common affiliations, venues, and fields of study. |
| fos | coauthor | Two fields of study are similar if they have papers with common authors. |
| fos | copaper | Two fields of study are similar if they appear in the same paper. |
| fos | covenue | Two fields of study are similar if they have papers from similar venues. |
| fos | metapath | Two fields of study are similar if they co-occur with common affiliations, venues, and fields of study. |
| journal | coauthor | Two journals are similar if they publish papers with common authors. |
| journal | cofos | Two journals are similar if they publish papers with similar fields of study. |
| journal | metapath | Two journals are similar if they co-occur with common affiliations, venues, and fields of study. |
Sample
Follow these samples for detailed usage information.
Reference
PySpark version
U-SQL version