Automated analysis of free speech predicts psychosis onset in high-risk youths Schizophrenia

GB contributed to the conception of the study, the interpretation of data, and drafting the manuscript. CMC led the prospective clinical high-risk cohort study and oversaw all data collection, and worked on and edited iterative drafts of the manuscript. DCJ also contributed to the design and conduction of the cohort study, and contributed suggested edits to the manuscript. FC and GAC designed and performed the automated text analysis; DFS and MS contributed to the analysis of the data; NM, SR, and MC collected and preprocessed data on patients with schizophrenia and their controls.

Content created with:

And so, we dedicate hours and hours every week to creating personas, hypothesising about audiences, segmenting users and running lengthy A/B tests to find the piece of content that our audience love. We add to our already-complex marketing stacks tools that tell us what messaging has been more successful, in order for us to optimise. In the climate of the current ‘data boom’, audience targeting naturally takes precedence, with the majority (55%) of marketers saying ‘better use of data’ for audience targeting is their priority in 2019, according to Econsultancy.

Associations between speech features and symptoms assessed with standard diagnostic instruments

By embracing semantic content analysis and working collaboratively with AI, we can feel confident in understanding exactly what content is going to work before we hit send. With time this company may be able to nail the truly useful features here and jettison some of what’s weighing the system down. Making powerful use of attention data in an RSS reader, without taking too much control away from readers, is a huge challenge. Offering extensive analytics in a clear fashion so the user experience isn’t too unpleasant is where Wizag has missed the boat so far. A river of news option to display individual items in order of publication across all feeds is a must have for me and Wizag is lacking that to date, but there’s more to life than chronology. RSS is world changing technology, but it risks getting bogged down in information overload.

An additional participant’s transcript was not included in speech analyses because her clinical outcome was indeterminate; she remained psychosis-free over 1.5 years of follow-up, but may have subsequently developed psychosis after the study. Findings support the utility of automated speech analysis to measure subtle, clinically relevant mental state changes in emergent psychosis. Recent developments in computer science, including natural language processing, could provide the foundation for future development of objective clinical tests for psychiatry. In this proof-of-principle study, our aim was to test automated speech analyses combined with Machine Learning to predict later psychosis onset in youths at clinical high-risk (CHR) for psychosis. Of note, the sample size employed in this initial study was small, with five participants developing psychosis during the follow-up period. This limitation meant that we were unable to divide participants into separate training and test samples, instead using cross-validation procedures to assess the predictive algorithm.

And, luckily, the ability to see what indisputably resonates the most with our audience – and drives our bottom-line – is already in our hands.

If we have just a few campaigns on the go, content analysis is easier, but it gets harder as we scale. It stops being practical to expect humans to spend days, weeks, even months labelling what goes into each piece of content. With this process, we can see which types of content are receiving the most engagement. And we can use these features to keep creating great campaigns that we further optimise as our understanding of customer content preferences grows.

• Drs Corcoran and Cecchi had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
• It stops being practical to expect humans to spend days, weeks, even months labelling what goes into each piece of content.
• Because the concept of semantic coherence we employed does not have a mathematical definition, in this validation we tested the coherence measure against a corpus of classic literature and assessed how the measure changed when we modified the original texts in a way that is relevant to the concept of semantic coherence.
• Schizophrenia, although relatively rare (lifetime prevalence ~1%), is among the most catastrophic mental illnesses both personally and societally.
• With this information, we trained the classifier to learn the features that discriminated among participants who did not subsequently develop psychosis (CHR−) from the group who did (CHR+).

Although clinicians routinely detect disorganized speech on the basis of clinical observations, our data suggest that automated analytic methods allow for superior assessment. As a direct, objective measure, automated speech analysis could thus provide important information to complement existing methods for patient assessment. Finally, these findings support the use of advanced computational methods to characterize complex human behaviors such as speech in both normal and pathological states. Such a fine-grained behavioral analysis could allow tighter mapping between psychiatrically relevant phenotypes and their underlying biology, in essence carving nature more closely at its joints. Better mapping between the behavioral and the biological is likely to lead to greater understanding of the pathophysiology of schizophrenia and other psychiatric disorders, potentially also informing psychiatric nosology.

Speech features were fed into a convex hull classification algorithm with leave-one-subject-out cross-validation to assess their predictive value for psychosis outcome. The canonical correlation between the speech features and prodromal symptom ratings was computed. Our findings from this proof-of-concept study, although needing to be replicated in larger samples, have several implications. First, reliable identification of individuals likely to progress to schizophrenia would greatly facilitate targeted early intervention.

The coherence measure as an index of ‘disorder’ in texts

Here, we present a proof-of-principle test of automated speech analysis to predict, at the level of the individual, the later onset of psychosis. Thirty-four CHR youths (11 females) had baseline interviews and were assessed quarterly for up to 2.5 years; five transitioned to psychosis. Using automated analysis, transcripts of interviews were evaluated for semantic and syntactic features predicting later psychosis onset.

Materials and methods

To complement the semantic analysis, we defined another measure for processing the documents, on the basis of Part Of Speech tagging (POS-Tag). For example, the sentence ‘The cat is under the table’ is tagged by the POS-Tag procedure as ((‘The’, ‘DT’), (‘cat’, ‘NN’), (‘is’, ‘VBZ’), (‘under’, ‘IN’), (‘the’, ‘DT’), (‘table’, ‘NN’)) where DT is the tag for determiners, NN for nouns, VBZ for verbs, and IN for prepositions. For every transcript, we calculated the POS-Tag information (with NLTK5) and used the frequencies of each tag as an additional attribute of the text. Tagging automation uses a hand-tagged corpus to train a parsing process using a variety of heuristics. The Trend Graph displays the fastest rising and fading topics in your feeds or all user feeds over whatever period of time you select.