A Compositional Model of Semantic Fluency

This repository contains the data and scripts for the paper: A Compositional Model of Semantic Fluency.

Authors: Surabhi S Nath, Alireza Modirshanechi, Peter Dayan

Abstract

The ability to recall semantically connected concepts---be it animals, summer fruits, or cities in Italy---is a remarkable capacity of the human mind. Such semantic fluency is thought to rely on traversing a mental space in which concepts are represented in terms of their meanings. However, the structure, properties, and navigability of this representational space remain enigmatic and highly debated. Existing approaches rely either on complex, uninterpretable distributional word-embeddings or on rigid, hand-crafted category norms. Here, we exploit the strengths of both, introducing Conceptome: a version of a compositional, interpretable, feature-based representation of semantic concepts, constructed by leveraging large language models. We use Conceptome to develop Conceptome-search, an auto-regressive model of how humans explore semantic spaces. We validate Conceptome and Conceptome-search using an animal fluency task, showing that they outperform state-of-the-art models in predicting human choices and capture key behavioral patterns such as interference. Our work, hence, offers new insights into the mechanisms underlying semantic fluency and memory retrieval. More broadly, our approach provides a general framework for constructing high-quality representations, with potential applications across cognition, including exploration, navigation, and creative thinking.

Repository Description

1. csvs contain the csv data files
2. figures contain the final figures used in the paper
3. files contain all config and auxillary files used in the code
4. fits contain the fit pickle files
5. models contain the model classes scripts and the main runner script
6. plots contain all plots for each figure, which were used to make the figures
7. scripts contain analysis scripts
8. simulations contain the simulation pickle files

Setup

We recommend setting up a python virtual environment and installing all the requirements. Please follow these steps:

git clone https://github.com/surabhisnath/Compositional_Fluency.git
cd Compositional_Fluency

python3 -m venv .env

# On macOS/Linux
source .env/bin/activate
# On Windows
.env\Scripts\activate

pip install -r requirements.txt

Running the code

NOTE: GPU required

models/runner.py is the main runner script for the following analyses. All settings can be set using arguments of runner.py. Before you being, ensure all models you wish to run are set to 1 in files/modeltorun.json.

1. To replicate Figure 1C: run cd scripts/Main; python make_TSNE.py. To plot the histogram of feature categories in Figure 1B: run cd scripts/main; python plot_feature_categories.py. All plots will be saved in plots/Figure1/

2. To analyse features and replicate Figure 2: run the notebook at scripts/Main/Model-free_Analysis.ipynb. All plots will be saved in plots/Figure2/

3. To replicate Figure 3: run cd scripts/Main; python model_NLLs.py to plot Figure 3B, and run cd scripts/Main/; python model_BLEUs.py to plot Figure 3C. model_nlls.png and model_bleus.png will be saved in plots/Figure3/

   To re-run model fitting and simulation: run cd models; python runner.py --fit --simulate (run overnight)

4. To replicate Figure 4: First perform feature ablation by running: cd models; python runner.py --ablation. Ablations saved as fits/ablations/ablations_Activity.pk and fits/ablations/ablations_HS.pk Then plot the figure: cd models; python runner.py --visweights. Plot saved as plots/Figure4/visweights.png

5. To replicate RT modelling in Figure 5 and Figure 6: Run: cd models; python runner.py --RT_analysis.

6. To replicate Figure 6 plots: Run: cd models; python runner.py --ARS. Plots saved as plots/Figure6/meanlogRT_transitions.png and plots/Figure6/meanprob_transitions.png

Citation

If you found this work useful, please consider citing us:

@article{nathcompositional,
  title={A Compositional Model of Semantic Fluency},
  author={Nath, Surabhi S and Modirshanechi, Alireza and Dayan, Peter},
  year={2026},
  publisher={OSF},
  url={https://osf.io/preprints/psyarxiv/adwzp_v1}, 
}