Linguamatics I2E takes NLP Text Mining to New Heights
Last week I was at the Linguamatics Text Mining Summit – for a feast of new experiences. The Summit was hosted at Wentworth-by-the-Sea, a new venue for the TMS, and it was a wonderful showcase for 3 days of workshops, round table discussions, and talks.
In addition to a new location, the attendees learned about fresh ways of approaching challenges across both life science and healthcare. Some of the innovations came from the recently released features of I2E 5.2, many others from our customers.
A relatively new use of Natural Language Processing (a well-established AI technology), is to power machine learning (ML). David Milward (Linguamatics) discussed how I2E both utilises ML and can also effectively feed ML workflows with high quality data. Simon Beaulah (Linguamatics) gave an overview of new applications of I2E NLP in healthcare; several of which involve using NLP to fuel ML models. These include predicting 30-day readmissions; extraction of cardiac risk factors; or patient stratification of heart failure risk from echocardiogram metrics.
A broad range of NLP Text Analytics Applications across Life Science and Healthcare
Using text mining for ETL (extract transform and load) is becoming more widespread. Several of our customers talked about the power of NLP to extract structured data from unstructured text, and load the results into databases, warehouses or data lakes for broader access and decision support:
- Dongyu Liu (Sanofi) uses I2E to extract associations with a broad range of HLA alleles and haplotypes with multiple sclerosis. Text mining has helped him to reveal significantly more annotations than previously compiled, and feed these into a database for use across discovery.
- Peter Hornbeck talked about the use of I2E at Cell Signaling Technology (CST). CST use I2E to extract key facts around cell communication from full text papers to curate and load into the Phosphosite database. Peter shared that since 2010, I2E has been invaluable at CST. He said “We would not be able to do what we do now without I2E – it is our saviour”.
Snippets from other customer presentations…
- Lue-Yen Tucker (Kaiser Permanente) talked on using I2E to extract patient data (uterine weights) from EHRs (pathology reports, operating notes). She said “Natural language processing extracted data faster and in a more replicable manner than manual extraction or using only regular expressions”.
- David Birtwell (Penn Medicine) also uses I2E for structuring EHRdata, and talked about the big benefits gained with I2E for improved phenotype identification for biobanking.
- Bryan Morganti (Pfizer) said that they had received “100% positive feedback from business teams” for implementing I2E within QA processes for regulatory applications.
- Ross Martin (CRISP) gave us another new experience with a striking personal oration on why improving heathcare informatics is so important to improving overall patient care.
- Allen Murvine and Roger Gildersleeve (IMO) talked about the value of I2E used with their medical ontologies for capturing the exact meaning in clinician’s minds (clinical intent) as they care for patients
- And Tony Sheaffer (Informatica) took us through data generations, data degradation, and how “mastering” information can transform a data swamp to a valuable data lake
Putting Text Mining’s power directly into the hands of scientists and business users
Another important theme was broadening access to the power of I2E through both democratic and programmatic access. Linguamatics showcased I2E Web Portals, which provide end user business professionals and scientists with quick and easy access to robust I2E text mining capabilities.
I2E Web Portals provide simplified, customizable web access to text mining over enterprise data silos. The presentation included a customer use case for text mining Clinical Investigator Brochures for safety assessment.
Linguamatics also gave the attendees a sneak peek into our exciting new development for next generation text mining, that includes a revolutionary “question routing engine” to get users rapidly and easily to the strongest answer to their search intent. Keep an eye on this space over the coming months for more details!