Top 20 TF-IDF Keywords from Dark Web CSAM Content Analysis

2,414 dark web pages were collected via a custom crawler from the Torch search engine for a comparative analysis of text mining techniques. Jin Gyeong Kim published this dataset on figshare in May 2026, which contains the top 20 keywords extracted using the TF-IDF method. The work evaluates TF-IDF, Eigenvector Centrality, and Word2Vec for extracting investigative keywords related to child sexual abuse materials.

Use Cases

• Benchmarking keyword extraction algorithms based on the comparative analysis of TF-IDF, Eigenvector Centrality, and Word2Vec described in the study
• Developing hybrid detection models based on the methodology combining graph-based ranking with semantic similarity
• Automating the expansion of early-stage crime indicators based on the dynamic keyword expansion approach proposed
• Improving operational efficiency in digital investigations based on the empirically collected, domain-specific dark web data

Strengths

• Dataset is derived from 2,414 empirically collected dark web pages, providing domain-specific context.
• The work presents a comparative evaluation of three distinct keyword extraction methods (TF-IDF, Eigenvector Centrality, Word2Vec).
• The proposed hybrid methodology was empirically tested for retrieval performance of CSAM-related sites.

Limitations

• The dataset is very small at 5.5 KB, indicating limited scope and likely only the top keyword results.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for training models.

Provenance

Source

Data collected from dark web pages indexed by the Torch search engine.

Collection Method

Collected via a custom crawler; keywords extracted using TF-IDF, Eigenvector Centrality, and Word2Vec methods.

Time Range

null

Freshness

Last updated 2026-05 08 17:23:26; freshness should be verified.

Geography

null