Spark NLP for Research and Education

You have our full support for using Spark NLP for Healthcare, Spark OCR, and the Data Library for open research & teaching projects

John Snow Labs

is making its licensed librariesfor state-of-the-art natural language processing – Spark NLP for Healthcare and Spark OCR – available under a free license for academic researchers, educators, and students. This includes over 1,000 pre-trained models as well as the entire catalog of over 2,220 expert-curated datasets in its Data Library. You can get a free personal license if you are doing academic research that will be publicly published under open-access, open-source, and open-data principles. If you are teaching a course that makes use of the library, you and your students can get a free license for it. We can also provide you with learning materials – from Python notebooks to slides & exercises. The free license includes the full capabilities of the software, all pre-trained models, and regular updates. Its goal is to enable you to easily reuse, reproduce, and improve production-grade, state-of-the-art NLP in your research & teaching. Fill in the form to apply for your free license. Please use your university’s email address and briefly explain how you will use the academic license. If you have more questions, feel free to reply to this email with your questions or proposed times for a call.

Sign up for a Free Academic License

Spark NLP for Healthcare

gives you access to state-of-the-art:

Clinical named entity

recognitionIngestion & Preparation – train your own or use pre-trained models to extract clinical facts (symptoms, diagnoses, treatments, procedures), drug facts (name, strength, dosage, route, frequency, duration), and biomedical terms (organism, tissue, gene, gene product, chemical, …).

Assertion status detection

telling between positive assertions (“patient has diabetes”), negative assertions (“no fever”), uncertain assertions (“shows indications of depression”), or assertions about other people (“family history of lung cancer”).

Entity resolution

train your own or use pre-trained models to resolve recognized entities to SNOMED-CT, ICD-10-CM, ICD-10-PCS, CPT, or RxNorm.

Relation extraction

use pre-trained models to automatically identify relations between entities such as drugs, dosage, duration, frequency, clinical events among many others.

Medical data normalization

normalize medications, lab results, vital signs, and demographic data – to simplify downstream analysis for extracted clinical information.

De-identification

Anonymize either structured tables or unstructured free text including all GDPR and HIPAA-required fields as well as and then either remove, mask, or obfuscate PHI.

Spark OCR

allows you to accurately transform PDF, DOCX, DICOM, and image files to digital text with built-in algorithms for:

image pre-processing (binarization, thresholding, erosion, scaling, skew correction)
image cleansing (noise scorer, remove objects, morphology), and
handling of complex document layouts (LayoutAnalyzer, SplitRegions, DrawRegions, PositionFinder).