The aim of Natural Language Processing, NLP for short, is to improve the communication between humans and machines. In order to get the most out of innovative and helpful digital solutions in the health care sector, the solutions need to be able to understand human language. However, this undertaking is not as easy as it sounds: human language comprises a lot of ambiguities and particularities that can lead to misunderstandings. This applies to the communication between humans, but also to the communication between humans and machines. Sarcasm, idioms, local dialects, homonyms, and homophones are just a few examples of natural language which can easily cause a misunderstanding. A simple change of region or lack of context can suddenly make it tough to understand the communication partner’s message in the right way. So, how can a machine understand our versatile language?
Core aspects and application of NLP
NLP turns the abundance of existing information into value. In other words: NLP makes data usable by analysing the language’s structure. One step in this process is “speech tagging”: the machine determines the grammar of a sentence, from which it identifies the subject, verb, and object, plus their dependencies. If a word has several meanings, it chooses the one fitting best with the identified context. Moreover, names are recognized and categorized, e.g., “Vienna” is identified as a location and “Mary” as a person. Also, coreference resolution and sentiment analysis are important steps for the machine in order to figure out the correct message. The first one means looking for words that refer to the same entity or concept – e.g., the words “Mary”, “she”, and “daughter” all can refer to the same person. Sentiment analysis makes it possible to read between the lines and decipher e.g., emotions and intentions of the sender of the message, as well as sarcasm.
The application areas of Natural Language Processing are diverse. Prominent examples from everyday life are voice assistants, email classification (important/spam), extracting relevant information from documents, automatic analyses of customer feedback, recruitment algorithms (screening applications), automatic text summaries, chatbots for legal and accounting professionals (searching for specific clauses), and disease diagnosis via a mix of health data and speech analysis.
Symptoma’s AI is trained for patient’s language
Symptoma is a digital health assistant. Patients and doctors visit symptoma.com, enter symptoms, answer questions, and receive a list of possible causes sorted by probability. Symptoma assists with difficult medical cases and helps uncover even ultra-rare diseases – because every patient deserves the right diagnosis and treatment.
Symptoma’s AI is trained to comprehend the patient’s natural language to support identifying the cause for their symptoms, and consequently, the appropriate therapy. This is a crucial part of the process, as patients often find it easier to talk about health problems when using their own words. The chatbot allows free text input; patients can even enter simple catchwords like “tiramisu”. Symptoma’s technology is able to put in into the right context and will in this case suggest “salmonellae” as cause. Consequently, Symptoma’s NLP contributes significantly to its extraordinary high diagnostic accuracy, helping millions of doctors and patients understand what is wrong. On top of that, Symptoma’s AI never stops learning: it is constantly getting smarter, as it learns from the continuously growing database.
Symptoma’s role in PERSIST
Symptoma’s role in PERSIST is to help reveal the huge potential of big data technology for improving the care delivery for cancer survivors. For this purpose, it analyses Electronic Health Records (EHRs), symptoms, signs, and risk factors of patients collected by the chatbot. As every disease and patient history is different, it is important to catch all data and put it in the right perspective in order to come to the best conclusion for the individual patient. A different context can change the whole story – one detail can be decisive. Therefore, Symptoma’s technology helps PERSIST to realise its goal to help cancer survivors and demonstrates the significance of digital support in the health care sector. Besides, it brings the Symptoma team closer to its aim: to enable precision medicine and eliminate medical guesswork.
Keywords: Natural Language Processing, NLP, chatbot, symptoms, symptom checker, precision medicine
Author’s name and position:
Annemarie Wiesner, Symptoma Marketing and Communications Manager
Alistair Tiefenbacher, Symptoma Chief Data Scientist
Company: Symptoma GmbH
Symptoma (@symptoma) / Twitter