Linked Crunchbase: A Linked Data API and RDF Data Set About Innovative Companies

Michael FärberarXiv 2019 · arXiv

Abstract

Crunchbase is a premier source of data on startups, investments, and tech personnel. However, its native JSON API limits complex querying and data integration. This paper introduces a Linked Data API that wraps Crunchbase into machine-readable RDF formats, alongside a methodology to crawl a massive 347-million-triple Crunchbase Knowledge Graph. This approach enables advanced SPARQL querying and seamless integration with the Linked Open Data (LOD) cloud.

Introduction

Crunchbase serves as a continuously updated, community-driven database containing unique information about innovative companies, funding rounds, acquisitions, and key personnel. Despite its richness, accessing this data for complex research or large-scale data integration remains challenging. The official Crunchbase API provides JSON output, which is suitable for simple lookups but inadequate for multi-hop queries or semantic integration. Previous attempts at Semantic Web wrappers have often failed due to API updates. Without an RDF representation, researchers cannot easily link Crunchbase entities to external knowledge bases like DBpedia or apply standard natural language processing (NLP) and data mining tools.

Methodology

We developed a Linked Data API wrapper that interfaces with the official Crunchbase REST API (v3.1). The architecture follows a multi-stage transformation process to ensure data integrity and semantic compatibility.

Transformation Workflow

Upon request, the wrapper fetches JSON payloads and transforms them into standard Semantic Web formats such as JSON-LD or N-Triples. The process is summarized in the following workflow:

Workflow for transforming the Crunchbase JSON API into a Linked Data API and RDF dataset.

Implementation Details

To maintain URI cleanliness and compliance with Linked Data best practices, API keys are passed via HTTP headers rather than URL parameters. The native Crunchbase schema is mapped to standard schema.org vocabularies (e.g., mapping a Crunchbase Person to schema:Person). This mapping ensures that the data is immediately usable by existing semantic tools.

The Linked Crunchbase Knowledge Graph

Beyond the live API, we established a methodology to build a complete RDF dump. By using Crunchbase's daily CSV exports as seed URIs and systematically crawling the wrapper API, we generated a knowledge graph containing approximately 347.7 million unique triples.

Ontology and Entity Relationships

The resulting graph captures the intricate web of relationships inherent in the startup ecosystem, as shown in the core ontology diagram below:

The core ontology of the Linked Crunchbase knowledge graph, illustrating relationships between organizations, people, and investments.

Entity Linking

The system automatically generates owl:sameAs links to DBpedia. To ensure high precision, we avoid naive string matching. Instead, we use strict criteria: Fully Qualified Domain Names (FQDNs) for organizations, and a combination of name and birthdate for individuals. This rigorous approach minimizes false positives in the linking process.

Results and Evaluation

The resulting dataset includes 16 entity types, covering 658k organizations, 780k people, 342k investments, and 222k funding rounds.

  • Precision: Manual evaluation demonstrated 89% precision for organization mappings and 95% precision for people mappings.
  • Notability Bias: The mapping process revealed a significant discrepancy: only 1,155 out of 659k organizations could be mapped to DBpedia. This highlights the "notability bias" of Wikipedia and underscores Crunchbase's value as a specialized knowledge graph for the business world.
  • Query Capabilities: The RDF format enables complex SPARQL queries, such as aggregating CEO demographics by continent, which would be prohibitively difficult using the standard JSON API.

Conclusion and Future Work

The Linked Crunchbase API achieves a 5-star Open Data deployment rating by utilizing standard vocabularies and linking to external schemas. While commercial licensing prevents direct distribution of the full 347-million-triple dump, the provided open-source scripts allow researchers to recreate it. Future work will focus on expanding links to Wikidata and utilizing the graph for temporal analysis of news articles to track the evolution of the tech landscape.

Summary updated

Abstract

Crunchbase is an online platform collecting information about startups and technology companies, including attributes and relations of companies, people, and investments. Data contained in Crunchbase is, to a large extent, not available elsewhere, making Crunchbase to a unique data source. In this paper, we present how to bring Crunchbase to the Web of Data so that its data can be used in the machine-readable RDF format by anyone on the Web. First, we give insights into how we developed and hosted a Linked Data API for Crunchbase and how sameAs links to other data sources are integrated. Then, we present our method for crawling RDF data based on this API to build a custom Crunchbase RDF knowledge graph. We created an RDF data set with over 347 million triples, including 781k people, 659k organizations, and 343k investments. Our Crunchbase Linked Data API is available online at http://linked-crunchbase.org.
BibTeX
@misc{farber2019linked,
  author = {Michael Färber},
  title = {{Linked Crunchbase: A Linked Data API and RDF Data Set About Innovative Companies}},
  year = {2019},
  eprint = {1907.08671},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1907.08671},
}
Figure