Improving spaCy dependency annotation and PoS tagging web service using independent NER services.

Area of Science:

• Computational Linguistics
• Natural Language Processing
• Bioinformatics

Background:

• Dependency parsing is crucial for text analysis but struggles with specialized terminology.
• Complex domain-specific language hinders the accuracy and efficiency of standard parsing methods.

Purpose of the Study:

• To investigate if named entity annotations enhance dependency parsing performance.
• To develop a service that leverages named entity recognition (NER) for improved parsing.

Main Methods:

• Developed a web service (BLAH5) integrating third-party named entity annotations.
• Evaluated the impact of NER on dependency parse quality and processing speed.

Main Results:

• Dependency parsing quality showed significant improvement when incorporating named entity annotations.
• Processing speed was also enhanced by the inclusion of NER data.

Conclusions:

• Named entity annotations are effective in boosting both the accuracy and speed of dependency parsing.
• The BLAH5 service provides a practical solution for improving text analysis in specialized domains.