A language independent acronym extraction from biomedical texts with hidden Markov models

Area of Science:

• Computational linguistics
• Bioinformatics
• Natural Language Processing

Background:

• Extracting acronyms and their definitions from unstructured text is challenging due to ambiguity and noise.
• Existing methods struggle with the complexities of acronyms in specialized domains like biomedicine.

Purpose of the Study:

• To propose and evaluate a novel probabilistic model for acronym and definition extraction.
• To address the challenges of ambiguity and noise in acronym identification using Hidden Markov Models (HMM).

Main Methods:

• Modeling the extraction process as a stochastic process using Hidden Markov Models (HMM).
• Utilizing characters of the acronym as states and text tokens as signals and observations.
• Calculating the most probable definition based on HMM output.

Main Results:

• Achieved high precision (93.50%) and recall (85.50%) in extracting acronyms and definitions.
• Demonstrated superior performance with the highest F1 score (89.40%) compared to other methods on a biomedical corpus.
• Successfully handled ambiguous and noisy acronym coinage.

Conclusions:

• The proposed HMM-based approach offers a robust solution for acronym extraction in challenging environments.
• Probabilistic modeling is effective in managing the inherent ambiguity and noise in acronym definition identification.
• This method shows significant potential for applications in biomedical text mining and information retrieval.