Per-topic term-frequency CSVs (Neurosynth v4 100-topic LDA)

Overview

One CSV per topic from the Neurosynth v4 100-topic LDA model, used as the "document" inputs to the OpenAI text-embedding-3-large embedding pipeline (see ../scripts/generate_topic_embeddings.py and ../scripts/codex_embedding_instructions.txt). Each CSV is named <topic_id>_<short_label>.csv, where the integer topic_id matches the topic index in v4-topics-100_<id>_<terms>_* NIfTI filenames in ../2016_Neurosynth_100_topics/. The file contains a single Term column listing the ~40 highest- probability terms for that topic, one per row.

Only a subset of the 100 LDA topics survives the CANlab curation step (topics dominated by methods / nuisance terms are dropped), so the filenames here are sparse over [0, 100).

This folder is CSV-only — no images, no MATLAB scripts.

Primary reference

Derived from the Neurosynth v4 LDA-100 topic model as part of the CANlab build pipeline; the underlying corpus and topic model are described in Yarkoni et al. 2011 Nat Methods.

How to use

Read any one CSV with MATLAB's readtable or Python's pandas:

T = readtable(which('60_Emotion_Processing.csv'));   % 1-column 'Term' table
top_terms = T.Term;

import pandas as pd
terms = pd.read_csv('60_Emotion_Processing.csv')['Term'].tolist()

Each topic's CSV becomes a "document" that is fed to OpenAI's text-embedding-3-large endpoint; the resulting vectors are saved into ../topic_embeddings_text-embedding-3-large.csv and the matching MAT-file.

File inventory

49 CSVs of the form <id>_<Label>.csv. The numeric prefix matches the topic index in the v4-100 model and in the matching NIfTI filenames. Representative examples:

File	Topic
0_Sensory_Stimulation.csv	Sensory / somatosensory stimulation, TMS, tDCS.
13_Executive_Function.csv	Executive control / cognitive control.
60_Emotion_Processing.csv	Emotion processing.
61_Pain_Perception.csv	Pain perception.
68_Working_Memory.csv	Working memory.
97_Fear_conditioning.csv	Fear conditioning.
99_Reward_processing.csv	Reward processing.

Run ls Neurosynth_maps/topic_terms_csv for the full list.

Citations

• Yarkoni T, Poldrack RA, Nichols TE, Van Essen DC, Wager TD (2011). Large-scale automated synthesis of human functional neuroimaging data. Nat Methods 8:665–670. doi:10.1038/nmeth.1635
• Poldrack RA, Mumford JA, Schonberg T, Kalar D, Barman B, Yarkoni T (2012). Discovering relations between mind, brain, and mental disorders using topic mapping. PLoS Comput Biol 8:e1002707. doi:10.1371/journal.pcbi.1002707
• OpenAI text-embedding-3-large model — see platform.openai.com/docs/guides/embeddings.