---

---

SocialED

A Python Library for Social Event Detection

The field of Social Event Detection represents a pivotal area of research within the broader domains of artificial intelligence and natural language processing. Its objective is the automated identification and analysis of events from social media platforms such as Twitter and Facebook. Such events encompass a wide range of occurrences, including natural disasters and viral phenomena.

SocialED is a comprehensive, open-source Python library designed to support social event detection (SED) tasks, integrating 19 detection algorithms and 14 diverse datasets. It provides a unified API with detailed documentation, offering researchers and practitioners a complete solution for event detection in social media. The library is built with modularity in mind, enabling users to adapt and extend components for various usages easily. SocialED supports a wide range of preprocessing techniques, such as graph construction and tokenization, and includes standardized interfaces for training models and making predictions. With its integration of popular deep learning frameworks, SocialED ensures high efficiency and scalability across CPU and GPU environments. Built adhering to high code quality standards, including unit testing, continuous integration, and code coverage, SocialED ensures robust, maintainable software.

Key Features

• Comprehensive Algorithm Collection: Integrates 19 detection algorithms and supports 15 widely-used datasets, with continuous updates to include emerging methods

• Unified API Design: Implements algorithms with a consistent interface, allowing seamless data preparation and integration across all models

• Modular Components: Provides customizable components for each algorithm, enabling users to adjust models to specific needs

• Rich Utility Functions: Offers tools designed to simplify the construction of social event detection workflows

• Robust Implementation: Includes comprehensive documentation, examples, unit tests, and maintainability features

SocialED includes 19 social event detection algorithms. For consistency and accessibility, SocialED is developed on top of DGL and PyTorch, and follows the API design of PyOD and PyGOD. See examples below for detecting outliers with SocialED in 7 lines!

SocialED plays a crucial role in various downstream applications, including:

• Crisis management

• Public opinion monitoring

• Fake news detection

• And more…

Social Event Detection Using SocialED with 5 Lines of Code:

from SocialED.dataset import Event2012                 # Load the dataset
dataset = Event2012()                                  # Load "Event2012" dataset

from SocialED.detector import KPGNN                    # Import KPGNN model
kpgnn = KPGNN(dataset, batch_size=200)                # Initialize KPGNN model

kpgnn.preprocess()                                     # Preprocess data
kpgnn.fit()                                           # Train the model
pres, trus = kpgnn.detection()                        # Detect events

kpgnn.evaluate(pres, trus)                            # Evaluate detection results

---

Implemented Algorithms

Algorithm	Year	Backbone	Scenario	Supervision	Ref
LDA	2003	Topic	Offline	Unsupervised	SocialED.detector.LDA
BiLSTM	2005	Deep learning	Offline	Supervised	SocialED.detector.BiLSTM
Word2Vec	2013	Word embeddings	Offline	Unsupervised	SocialED.detector.Word2Vec
GloVe	2014	Word embeddings	Offline	Unsupervised	SocialED.detector.GloVe
WMD	2015	Similarity	Offline	Unsupervised	SocialED.detector.WMD
BERT	2018	PLMs	Offline	Unsupervised	SocialED.detector.BERT
SBERT	2019	PLMs	Offline	Unsupervised	SocialED.detector.SBERT
EventX	2020	Community	Offline	Unsupervised	SocialED.detector.EventX
CLKD	2021	GNNs	Online	Supervised	SocialED.detector.CLKD
KPGNN	2021	GNNs	Online	Supervised	SocialED.detector.KPGNN
FinEvent	2022	GNNs	Online	Supervised	SocialED.detector.FinEvent
QSGNN	2022	GNNs	Online	Supervised	SocialED.detector.QSGNN
ETGNN	2023	GNNs	Offline	Supervised	SocialED.detector.ETGNN
HCRC	2023	GNNs	Online	Unsupervised	SocialED.detector.HCRC
UCLSED	2023	GNNs	Offline	Supervised	SocialED.detector.UCLSED
RPLMSED	2024	PLMs	Online	Supervised	SocialED.detector.RPLMSED
HISEvent	2024	Community	Online	Unsupervised	SocialED.detector.HISEvent
ADPSEMEvent	2024	Community	Online	Unsupervised	SocialED.detector.ADPSEMEvent
HyperSED	2025	Community	Online	Unsupervised	SocialED.detector.HyperSED

Modular Design and Utility Functions

SocialED is built with a modular design to improve reusability and reduce redundancy. It organizes social event detection into distinct modules:

• preprocessing

• modeling

• evaluation

The library provides several utility functions including:

• utils.tokenize_text and utils.construct_graph for data preprocessing

• metric for evaluation metrics

Library Robustness and Accessibility

Quality and Reliability

• Built with robustness and high-quality standards

• Continuous integration through GitHub Actions

• Automated testing across Python versions and operating systems

• >99% code coverage

• PyPI-compatible and PEP 625 compliant

• Follows PEP 8 style guide

Accessibility and Community Support

• Detailed API documentation on Read the Docs

• Step-by-step guides and tutorials

• Intuitive API design inspired by scikit-learn

• Open-source project hosted on GitHub

• Easy issue-reporting mechanism

• Clear contribution guidelines

---