🧵 1/ Excited to share my latest #research Our paper, "Approximating the semantic space: word embedding techniques in psychiatric speech analysis", is now published in @schizjournal by @SpringerNature. Read it here:

En mi curso Fundamentos de Product Management, aprendemos a abordar la priorización de forma práctica y estratégica. Estaré dando charlas gratuitas via (Zoom) los días 21, 22 y 23 de octubre a las 19:00 (Chile). 🔗 Link para inscribirte en las charlas acá

Gaslight (George Cukor, 1944)

I want to thank our entire team and collaborators who have made this work possible!

3. A foundation for future multilingual, cross-cultural, and neurobiologically-informed research in psychiatric linguistics.

2. A way to monitor symptom progression and remission using naturalistic speech.

This work provides: 1. A unified, quantifiable tool for characterizing language disruptions in psychosis.

📈 Methodology Speech samples were analyzed using static (fastText) and contextualized (BERT, SentenceTransformers) language models. Semantic similarities between words and sentences were computed and reduced via PCA to generate the final composite index.

🧠 Symptom tracking The index not only differentiated between acute and remitted states but also predicted 65% of the variance in TLC scores, linking linguistic behavior to clinical symptoms like derailment and loss of goal.

Key findings: 🔍 Composite semantic index A single, agnostic semantic index successfully captured meaningful differences in speech patterns, distinguishing SSD from HC and tracking symptom changes longitudinally.

Our goal was to assess whether this index could (i) distinguish individuals with schizophrenia spectrum disorders (SSD) from healthy controls (HC), (ii) reflect changes from acute psychosis to remission, and (iii) predict clinical language disturbances measured by the TLC scale.

Using principal component analysis (PCA) on 117 semantic features derived from different language models, we created an interpretable measure that tracks linguistic disturbances over time.

In this study, we developed a single composite semantic index to summarize the high-dimensional semantic space extracted from spontaneous speech in psychosis.

Disorganized speech is a hallmark of psychotic disorders, yet capturing its linguistic disruptions systematically has been a challenge.

I’m happy to share that our publication “A single composite index of semantic behavior tracks symptoms of psychosis and depression across diagnostic boundaries” is now available in Schizophrenia Research. You can read the full article here 👀 https://t.co/QiHUJLb4iI

Este proyecto agarró vuelo porque está siendo popular en TikTok. En las últimas 48 horas hemos recibido más de 4000 consultas nuevas en el sitio. ¿Cuáles son los términos más buscados para explorar la historia de nuestro humor? Nóctulo, suegra, pico, pene y piñera😅

New study exploring how word embedding techniques can be applied to analyse speech patterns in psychiatric disorders, focusing on the schizophrenia spectrum and major depression @PossiblyHuman_ @schizjournal @SIRSGlobal

Linguistics is everywhere! Who would have thought our field would turn out to be so popular? Law, biology, machine learning, they all want to know about linguistics now. 101 in Language and AI in 101s: https://t.co/7enIqE1OJQ

🚨 New Preprint!! LLMs trained on next-word prediction (NWP) show high alignment with brain recordings. But what drives this alignment—linguistic structure or world knowledge? And how does this alignment evolve during training? Our new paper explores these questions. 👇🧵

Honestly difficult to imagine anything more lazy and midwit than this anodyne take about outsourcing your thinking to a black box statistical algorithm. You guys think you're being so trailblazing and contrarian for repeating the same marketing copy over and over? Touch grass

🗣️Language is widely distributed throughout the brain🧠 In a recent correspondence in @NatRevNeurosci, we suggest that there is no 'language network' in the brain. What appears as such is an inevitable illusion created in part by the methods we use. https://t.co/TCyEQCKcLJ