PoTeC - Potsdam Textbook Corpus

This repository contains the Potsdam Textbook Corpus (PoTeC) which is a natural reading eye-tracking corpus. Four groups of participants (expert/beginner level students of physics and biology) read 12 short texts taken from textbooks of physics and biology while their eye movements were monitored. The final dataset contains the reading data for 75 participants each reading all 12 texts. The study follows a 2x2x2 fully-crossed factorial design:

• Factor 1: Study discipline of participant with the levels either physics or biology
• Factor 2: Study level of participant with the levels either beginner or expert
• Factor 3: Text domain with the levels either physics or biology

	Physics	Biology
Beginner	12	16
Expert	20	27

Both factors are quasi-experimental and manipulated between subjects. The readers' text comprehension as well as their background knowledge on the topics presented in the texts were assessed by multiple-choice questions.

More information is found in the following README'S:

• preprocessing
• participants
• stimuli
• eye-tracking data
• additional processing

Download the data

The data files are stored in an OSF repository. If the repository has been cloned, they can be downloaded and extracted automatically using the following script:

# or python3
python download_data_files.py

# OR to extract the files directly
python download_data_files.py --extract

Alternatively, they can be downloaded manually from the OSF repository and extracted into the respective folders.

pymovements integration

PoTeC is integrated into the pymovements package. The package allows to easily download the raw data and further process it. The following code snippet shows how to download the data:

# pip install pymovements
import pymovements as pm

dataset = pm.Dataset('PoTeC', path='data/PoTeC')

dataset.download()

Note on reading in the data files

The German text p3 includes the word "null". If e.g. the word features are read using pandas, the word "null" is interpreted as a NA value. In order to avoid this behavior the command can be used with the following arguments:

import pandas as pd
pd.read_csv('word_features_p3.tsv', sep='\t',  
            keep_default_na=False,
            na_values=['#N/A', '#N/A N/A', '#NA', '-1.#IND', '-1.#QNAN', '-NaN', '-nan',
                       '1.#IND', '1.#QNAN', '<NA>', 'N/A', 'NA', 'NaN', 'None', 'n/a',
                       'nan', '']
            )

Data Overview

The data that was used to create the corpus and that was obtained during the experiments is made available in various stages. The data is stored in respective sub folders each of which contains a README that provides more information about the data and how to use it. For a detailed description of the data types, format and content, please refer to the CODEBOOK.

This repository contains the following data:

• Eye-tracking data
  • raw eye-tracking data
  • preprocessed eye-tracking data
• Stimuli
  • stimuli texts
  • text and background questions
• Anonymized participant data
• Scripts (in Python)
  • scripts to preprocess the data
  • additional scripts that have been used to process the data further

The scripts were run using Python 3.9 with the dependencies specified in the requirements.txt file.

Technical set-up

The experiment was run with the following technical set-up:

	Setting	Value
Technical set-up	Eye-tracking device	Eyelink 1000, dektop mounted camera system with a 35 mm lens
	Sampling rate	1000 Hz
	Monitor size	47.5x30 cm, 22 inch
	Monitor resolution	1680x1050 pixels
	Eye-to-screen distance	61 cm
	Eye-to-camera distance	65 cm
	Experiment software	Experiment Builder software provided by SR Research
Stimulus presentation	Background color	Black
	Font color	White
	Font size	18
	Font	Courier
	Stimulus size	On average 158 words shown on multiple lines on one page
	Number of characters per visual angle (middle of screen)	2.8 characters per degree of visual angle
	Spacing

Stimuli

Please note that the full stimuli texts are not yet available. Contact deborahnoemie.jakobi(at)uzh.ch for more information.

Stimuli Annotation

The stimuli have been manually annoted with part-of-speech tags and other linguistic information. The annotations are described in a separate file: ANNOTATION.

Citation

@misc{potec,
    url={\url{https://github.com/DiLi-Lab/PoTeC}},
    author={Jakobi, Deborah N. and Kern, Thomas and Reich, David R. and Haller, Patrick and J\"ager, Lena A.},
    title={{PoTeC}: A {German} Naturalistic Eye-tracking-while-reading Corpus},
    year={2024},
    note={under review}
}

Repository Structure

PoTeC-data
├── CODEBOOK.md
├── README.md
├── requirements.txt
├── additional_scripts
│   ├── ADDITIONAL_SCRIPTS.md
│   ├── compute_reading_measures.py
│   ├── generate_scanpaths.py
│   ├── merge_reading_measures.py
│   ├── create_codebook_tables.py
│   ├── surprisal.py
│   ├── get_surprisal.py
│   ├── merge_fixations_and_coordinates.py
│   ├── merge_scanpaths.py
│   ├── analyses.R
│   ├── run_bayesian_models.R
│   ├── run_freq_models.R
│   ├── all_colls_description.csv
│   └── all_codebook_texts.csv
├── eyetracking_data
│   ├── EYETRACKING_DATA.md
│   ├── original_uncorrected_fixation_report.txt
│   ├── fixations
│   │   └── ...
│   ├── fixations_uncorrected
│   │   └── ...
│   ├── asc_files
│   │   └── ...
│   ├── raw_data 
│   │   └── ...
│   ├── reader_merged
│   │   └── ...
│   ├── reading_measures
│   │   └── ...
│   ├── scanpaths
│   │   └── ...
│   └── scanpaths_merged
│       └── ...
├── participants
│   ├── PARTICIPANTS.md
│   └── participant_data.tsv
├── preprocessing_scripts
│   ├── PREPROCESSING_SCRIPTS.md
│   ├── char_index_to_word_index.py
│   ├── create_word_aoi_limits.py
│   ├── correct_fixations.py
│   ├── split_fixation_report.py
│   ├── asc_to_csv.py
│   ├── aoi_to_word.tsv
│   ├── sent_limits.json
│   └── word_limits.json
└── stimuli
    ├── ANNOTATION.md
    ├── STIMULI.md
    ├── practice_items.txt
    ├── dependency_trees_manually_corrected.tsv
    ├── aoi_texts
    │   └── ...
    ├── stimuli
        ├── stimuli.bib
    │   ├── items.tsv
    │   └── stimuli.tsv
    └── word_features
        └── ...