A top 5 global Pharma company needed to improve the process of scientific writing out of already available information. They wanted to create a smart knowledge discovery solution that would help their researchers find the relevant information they needed to analyze while creating a clinical trial synopsis document. This information was locked in huge volumes of data stored across various structured (proprietary or public registers) and unstructured (clinical study protocols and reports, etc.) sources. This made it extremely difficult to find the most relevant information in all related clinical documents that would help them prepare the scientific article draft.
As the existing process could not meet the needs of their clinical trial researchers, the Pharma company wanted to implement an AI-based solution that would help them:
One of the main challenges was the data variety. The data was stored in different types of documents (clinical protocols, TFL, scientific articles, etc.) and in various file formats (Text, PDF, Scanned images, etc.). What’s more, the information was often provided with different levels of detail. All this required a lot of manual effort to get to the relevant information, and there was always the possibility of missing something important.
As the information was scattered among different documents and was described with different granularity, researchers spent too much effort in collecting the required data points, analyzing the information and summarizing it into meaningful conclusions. Often, the information was ambiguous and duplicated, which additionally slowed down the process.
The overall authoring process had to ensure the generation of unique content out of the available source data. While most of the information was already present in the source documents and could be reused for scientific writing, it still required a certain level of uniqueness of the produced text while keeping the meaning intact.
The smart AI solution jointly developed by Ontotext and Wipro helped the Pharma company transform and enhance their scientific writing process. It consisted of automatic data extraction, definition of business rules and natural language generation.
Using advanced natural language processing pipelines, Ontotext extracted specific key categories such as introduction, method data (Study Design, Study Population, etc.) and result data (Patient Disposition, Patient Demographics, Safety, etc.). Then the clinical trial data extracted from the documents was populated into a custom built knowledge graph and interlinked with the Pharma company’s clinical trials public data. As a result, all the Pharma company’s data was semantically normalized to specific clinical concepts (treatment, conditions, etc) and could be used for automatic generation of human readable text.
On top of this high quality structured data, Wipro’s domain experts applied business rules associated with the medical information needs. They also reviewed the extracted content and put the resulting information into the context of the rest of the data.
Finally, on top of the knowledge graph, Ontotext applied data analytics techniques to extract the important facts and to generate meaningful natural language summarizations for each knowledge category. Now the Pharma company researchers could specify the context in which they wanted to filter down the huge volumes of existing documents to a manageable subset of relevant documents. This semantically normalized high granularity data was a powerful knowledge discovery enabler.
With the smart scientific writing solution developed by Ontotext and Wipro, the Pharma company researcher can now:
Do you think this case resembles your particular needs?