Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Triplètoile: Extraction of Knowledge from Microblogging Text

Vanni Zavarella, Sergio Consoli, Diego Reforgiato Recupero, Gianni Fenu, Simone Angioni, Davide Buscaldi, Danilo Dessì, Francesco Osborne

TL;DR

Triplétoile addresses the challenge of constructing open-domain knowledge graphs from noisy microblogging text by combining dependency-pattern extraction, unsupervised relation clustering, and a robust entity refinement pipeline. The approach yields a scalable architecture that processes ~100k tweets to produce a substantial knowledge base (DTSMM_KG) with 22,270 statements, achieving precision > 0.95 and outperforming several baselines by ~5% in precision while generating more triples. A Digital Transformation use case demonstrates practical utility, including extensive DBpedia grounding and a public data release, with potential extension to RAG-enabled downstream tasks. The work highlights both the feasibility and value of real-time, cross-source KG construction from social media, while outlining future directions such as domain ontologies, supervised relation mapping, and integration with large language models.

Abstract

Numerous methods and pipelines have recently emerged for the automatic extraction of knowledge graphs from documents such as scientific publications and patents. However, adapting these methods to incorporate alternative text sources like micro-blogging posts and news has proven challenging as they struggle to model open-domain entities and relations, typically found in these sources. In this paper, we propose an enhanced information extraction pipeline tailored to the extraction of a knowledge graph comprising open-domain entities from micro-blogging posts on social media platforms. Our pipeline leverages dependency parsing and classifies entity relations in an unsupervised manner through hierarchical clustering over word embeddings. We provide a use case on extracting semantic triples from a corpus of 100 thousand tweets about digital transformation and publicly release the generated knowledge graph. On the same dataset, we conduct two experimental evaluations, showing that the system produces triples with precision over 95% and outperforms similar pipelines of around 5% in terms of precision, while generating a comparatively higher number of triples.

Triplètoile: Extraction of Knowledge from Microblogging Text

TL;DR

Triplétoile addresses the challenge of constructing open-domain knowledge graphs from noisy microblogging text by combining dependency-pattern extraction, unsupervised relation clustering, and a robust entity refinement pipeline. The approach yields a scalable architecture that processes ~100k tweets to produce a substantial knowledge base (DTSMM_KG) with 22,270 statements, achieving precision > 0.95 and outperforming several baselines by ~5% in precision while generating more triples. A Digital Transformation use case demonstrates practical utility, including extensive DBpedia grounding and a public data release, with potential extension to RAG-enabled downstream tasks. The work highlights both the feasibility and value of real-time, cross-source KG construction from social media, while outlining future directions such as domain ontologies, supervised relation mapping, and integration with large language models.

Abstract

Numerous methods and pipelines have recently emerged for the automatic extraction of knowledge graphs from documents such as scientific publications and patents. However, adapting these methods to incorporate alternative text sources like micro-blogging posts and news has proven challenging as they struggle to model open-domain entities and relations, typically found in these sources. In this paper, we propose an enhanced information extraction pipeline tailored to the extraction of a knowledge graph comprising open-domain entities from micro-blogging posts on social media platforms. Our pipeline leverages dependency parsing and classifies entity relations in an unsupervised manner through hierarchical clustering over word embeddings. We provide a use case on extracting semantic triples from a corpus of 100 thousand tweets about digital transformation and publicly release the generated knowledge graph. On the same dataset, we conduct two experimental evaluations, showing that the system produces triples with precision over 95% and outperforms similar pipelines of around 5% in terms of precision, while generating a comparatively higher number of triples.
Paper Structure (21 sections, 3 equations, 6 figures, 6 tables)

This paper contains 21 sections, 3 equations, 6 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. The Proposed Architecture
  4. Data Preprocessing
  5. Triple Extraction
  6. Entities
  7. Relations
  8. Entity Refining
  9. Relation Refining
  10. Relation Embeddings
  11. Dimensionality Reduction and Clustering
  12. Relation Mapping
  13. The Use Case: Digital Transformation Monitoring
  14. Preprocessing and Triple Extraction
  15. Relation Clustering
  16. ...and 6 more sections

Figures (6)

  • Figure 1: Flowchart of the pipeline for generating a knowledge graph from micro-blogging text data.
  • Figure 2: Example of tweet preprocessing.
  • Figure 3: Visualization of candidate entities extracted from a sample of tweets.
  • Figure 4: A shortened example of reification for a Statement concerning the instance machine_learning, grounded by 6 tweets, with the three dots referring to the hidden dtsmm-ont:comesfromTweet predicates.
  • Figure 5: A subgraph from DTSMM_KG showing a few sample claims for the instance machine_learning, prior to applying explicit statement reification.
  • ...and 1 more figures