More than Words

This repository contains all the parsing and processing scripts used in the manuscript: "More than words: Can free reports adequately measure the richness of perception?". This work describes a series of three experiments run in two batches (exploratory, replication), where participants were exposed to briefly-presented images, and then asked to provide five words describing their experience.

All experiments, parsing and processing codes were programmed by Rony Hirschhorn.

Processing Codes

Data processing is done separately per experiment.

Main Experiments

Parsing

The processing module for the main experiments is managed by manage_analysis.py:

• The first stage is to pre-process all the raw data files into something coherent, and attribute each participant with all the responses they made during the experiment. This is done by calling the process_raw_data.py module's manager function (manage_preprocessing).
• Then, once all data is loaded from the files, the actual parsing, processing and descriptive stats can begin. This is done by the analyze_data.py module (by calling its manager function, manage_data_analysis).

NOTE: during word parsing, we use nltk's stop words (from nltk.corpus import stopwords). This requires installing nltk and downloading stop words:

• Either by following the import line with nltk.download('stopwords')
• Or straight from the commandline: python -m nltk.downloader stopwords

Aggregation

Once done processing each experiment separately, the results are then aggregated for plotting and more descriptives purposes. This is a separate process, that is managed by the aggregate_results.py module.

Additional Datasets

Furthermore, our experiments follow the design of this experiment. In case one is interested in stats about their data calculated in our scripts, the analyze_Chuyin_data.py module reshapes the data from this experiment to match the shape of our experimental outputs; then, it runs the manage_analysis module.

Control Experiments

In addition, two separate control experiments were performed in our study. To parse and aggregate their results, a module called control_analysis.py handles the processing of both experiments, separately.