Traditional NLP discards the language that contains the most important indicators for understanding mental health and measuring the therapeutic alliance.
Historically, NLP practitioners have removed "stop words" from analyses because stop words have less obvious value than nouns, verbs, and adjectives. But psychologists who study language have long recognized that the way that people use stop words reveals a great deal about the psychology of the person behind the language.
What are stop words?
Stop words are sometimes also called "function words,” and they are the words that convey the most human parts of language - pronouns, prepositions, articles, conjunctions, and auxiliary verbs. The way that people use these words provides a great deal of insight into their levels of stress and anxiety, their stability, and their sense of self.
While the typical vocabulary contains roughly 30,000 words, stop words account for less than 500 but make up more than half of the words we use in conversation. Interestingly, stop words are processed in the brain differently, and they are used largely unconsciously. But, they hold the key to understanding psychology, personality, relationships, and some of the most important information about the minds of people who use digital health platforms.
Measure mental health, the therapeutic alliance, and drivers of program adherence.
The Receptiviti API incorporates stop word analysis (SWA) and identifies language-based psychological indicators associated with mental health and distress-related concerns such as depression, stress, and social anxiety, and also enables measurement of the therapeutic alliance - one of the most important predictors of treatment success.
Our models have been constructed and extensively validated by panels of psychologists and have been cited in over 19,000 peer-reviewed research studies, many of which focus on understanding the linguistic fingerprints of mental health, coping, disease, recovery, and treatment delivery. And, in a recent study conducted by researchers at Leiden and Utrecht Universities, Receptiviti outperformed BERT, 95.2% vs. 60.7%, in predicting mental disorders from language.
Talk with one of our digital health specialists:
Have questions, or want to learn more about how digital health providers typically use Receptiviti in their research and products?
Provide your contact information and one of our digital health specialists will contact you within 24 hours:
Watch the TEDTalk
Receptiviti co-founder Dr. James Pennebaker discusses the fascinating relationship between stop words and human psychology.
Research from Leiden and Utrecht Universities finds Receptiviti’s science outperforms BERT and neural network based approaches in predicting mental disorders based on language.
Language markers for mental disorders can help to diagnose a person. Research thus far on language markers and the associated mental disorders has been done mainly with the Linguistic Inquiry and Word Count (LIWC) program. In order to improve on this research, we employed a range of Natural Language Processing (NLP) techniques using LIWC, spaCy, fastText and RobBERT to analyze Dutch psychiatric interview transcriptions with both rule-based and vector-based approaches. LIWC in combination with the random forest classification algorithm performed best in predicting whether a person had a mental disorder or not (accuracy: 0.952; Cohen’s kappa: 0.889). SpaCy in combination with random forest predicted best which particular mental disorder a patient had been diagnosed with (accuracy: 0.429; Cohen’s kappa: 0.304).
See Receptiviti API in action!
Receptiviti is conducting the world's largest continuous study of the pandemic's impact on the mental health of doctors and nurses.
This live index uses the Receptiviti API to passively measure the emotional and psychological health of over 6,500 doctors and nurses on a daily basis by analyzing language in their posts on Reddit -- every day since July 2019 through the current day.
See the index
Analyzing the language people use when they write, speak, and communicate with others, including their interactions with therapists and online mental health platforms, can provide insight into their psychological state and their status along their journey toward healing.
The Receptiviti API identifies many of the language-based psychological indicators associated with mental health and distress-related concerns such as depression, stress, and social anxiety.
Our models have been constructed and extensively validated by panels of psychologists and have been cited in over 19,000 research studies, of which over 6,500 focus on understanding the linguistic fingerprints of mental health, coping, disease, recovery, and treatment delivery.
How is Receptiviti used by
virtual therapy platforms?
Receptiviti helps virtual therapy platforms enhance their diagnostic capabilities, reduce patient attrition, and improve outcomes. Our indicators can be used to measure the therapeutic alliance between provider and patient, monitor provider empathy, identify signals of comorbidities, and track treatment progression and success:
Increase the likelihood of successful outcomes identifying the psychological markers of progression and success through changes in functional language patterns. For example, markers of reduced anxiety include diminishing use of words associated with negative emotion, anxiety, causation, and insight, and increasing use of past-tense related language.
Improve patient outcomes by measuring and supporting healthcare provider empathy skill development. According to Lambert and Barley (2001), 30% of patient outcomes in psychotherapy can be attributed to the therapeutic alliance and facilitative conditions, such as empathy, warmth, and congruence. Receptiviti measures empathy through analysis of provider language style, and through quantification of language synchrony between healthcare providers and patients.
Improve patient adherence by tailoring communications to each patient’s psychology and cognitive style. By analyzing patient language, Receptiviti uncovers how they think, make decisions, and interpret information, enabling providers to optimize communication style with patients and platform users to increase engagement, improve adherence, and increase the likelihood of treatment success.