Improved Chemical Text Mining of Patents
Author Information
Author(s): Sayle Roger A, Petrov Plamen, Winter Jon, Muresan Sorel
Primary Institution: NextMove Software
Hypothesis
The study investigates the effectiveness of automatic chemical text mining methods in extracting key compounds from pharmaceutical patents.
Conclusion
Novel algorithms for automatic chemical spelling correction significantly improve the extraction of key compounds from patents.
Supporting Evidence
- Traditional text mining methods struggle with the unique challenges posed by pharmaceutical patents.
- The study highlights the importance of quality over quantity in extracting chemical structures from patents.
- High rates of OCR errors and typos significantly hinder the effectiveness of traditional extraction methods.
Takeaway
This study shows that fixing spelling mistakes in patent data helps find important chemical information better.
Methodology
The study analyzed both a small hand-curated benchmark set and a large-scale database of 12 million patents using automatic Chemical Named Entity Recognition methods.
Limitations
The study is limited by the high rate of OCR errors and typos in patent-office data feeds.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website