eTBLAST

eTBLAST performs text-similarity searches to retrieve and rank biomedical literature (including MEDLINE) and to identify related journals, authors, and topical trends.

Key Features:

• Hybrid search system: Combines a fast, low-sensitivity weighted keyword-based retrieval with a sentence-alignment–based similarity algorithm for refined ranking.
• Sentence-alignment similarity: Uses a sentence-alignment based ranking algorithm to assess document-level similarity beyond keyword overlap.
• Weighted keyword retrieval: Employs a weighted keyword algorithm for rapid initial retrieval with broad recall.
• Abstract parsing and summarization: Parses and summarizes abstracts to support similarity detection and trend extraction.
• Performance characteristics: Aims to optimize precision and recall while enhancing sensitivity and computational speed via its hybrid approach.

Scientific Applications:

• Literature search and retrieval: Retrieve and rank related biomedical documents from databases such as MEDLINE based on textual similarity.
• Journal identification: Identify journals suitable for a given abstract by matching abstract similarity to journal literature.
• Expertise mapping: Detect authors with expertise in specific topics by analyzing similarity among author-associated abstracts.
• Trend analysis: Summarize search results to reveal topical trends and emerging research areas.

Methodology:

Hybrid pipeline combining weighted keyword-based retrieval and sentence-alignment–based similarity ranking, together with abstract parsing and summarization and the implementation of standard and novel text-similarity algorithms.