Recently @malaysianpaygap has gotten quite famous as a platform that enables workers throughout Malaysia to anonymously share their salaries amongst other Malaysians. Its a great initiative and I am fully supportive behind ensuring that Malaysians are not taken advantage of by companies and get a liveable wage(especially when inflation is sky high).
UPDATE: I am removing access to any of the data after consulting the MalaysianPayGap team as they are worried of people using the data to impart their own biases into their analysis.
- Run the following to get conda environment setup
conda create --name pay python=3.7
conda activate pay
pip install -r requirements.txt- Next we will need to scrape all the data from Instagram manually using BeautifulSoup! Just kidding I am too lazy so I will be using InstaLoader to do all the heavy lifting for me. The conda environment will have it installed for you already.
# you might need to pass in your username to login
instaloader --login=USERNAME malaysianpaygap --dirname-pattern={profile} --comments --no-profile-pic --post-metadata-txt="Caption: {caption}\n{likes} likes\n{comments} comments\n" --filename-pattern={date_utc:%Y}/{shortcode}This should create the following directory structure:
|-- malaysianpaygap
| |-- 2022
| | |-- CaRp-1uPh8l.jpg # image
| | |-- CaRp-1uPh8l.json.xz
| | |-- CaRp-1uPh8l.txt # text data which was specified under --post-metadata-txt
| | |-- CaRp-1uPh8l_comments.json # all the comments
| | |-- CaT5MguPpDI.jpg
| | |-- CaT5MguPpDI.json.xz
| |-- 2022-02-27_04-58-58_UTC_profile_pic.jpg
| |-- id
| `-- malaysianpaygap_47523401972.json.xz
|-- requirements.txt
|-- scripts
| `-- entrypoint.sh
`-- src
|-- __init__.py
|-- extract_text_images.py
|-- main.py
|-- preprocess_comments.py
`-- preprocess_images.py
NOTE: Please do NOT change the directory structure, it will break the entire pipeline.
- You should have everything ready to run the preprocessing scripts that I have made! I have a bash script that runs everything in the correct order.
# make bash script runnable
chmod +x scripts/entrypoint.sh
bash scripts/entrypoint.shYou should see the following output:
2022-03-02 22:59:54.012 | INFO | src.preprocess_comments:main_preprocess_comments:83 - Running preprocess_comments
2022-03-02 22:59:56.276 | INFO | src.preprocess_comments:main_preprocess_comments:110 - DataFrame saved to /Users/yravindranath/pay/data/comments.csv
2022-03-02 22:59:56.277 | INFO | src.preprocess_comments:main_preprocess_comments:111 - Completed preprocess_comments
2022-03-02 22:59:57.537 | INFO | src.preprocess_images:main_preprocess_images:140 - Running preprocess_images
2022-03-02 22:59:57.840 | INFO | src.preprocess_images:main_preprocess_images:160 - DataFrame saved to /Users/yravindranath/pay/data/posts.csv
2022-03-02 22:59:57.841 | INFO | src.preprocess_images:main_preprocess_images:161 - Completed preprocess_images
2022-03-02 22:59:59.099 | INFO | src.extract_text_images:main_extract_text_images:54 - Running extract_text_images
Pandas Apply: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 159/159 [02:09<00:00, 1.23it/s]
2022-03-02 23:02:25.087 | INFO | src.extract_text_images:main_extract_text_images:70 - DataFrame saved to /Users/yravindranath/pay/data/posts_text.csv
2022-03-02 23:02:25.088 | INFO | src.extract_text_images:main_extract_text_images:71 - Completed extract_text_images
A new directory data will be created like so:
|-- data
| |-- comments.csv
| |-- comments.json
| |-- posts.csv
| |-- posts_text.csv
| `-- processed_images
| |-- CaRp-1uPh8l.jpg
| |-- CaT5MguPpDI.jpg
| |-- CaT6d2Yve5X.jpg
In the next section I will go over the data that was created.
comments.csv - Contains all the comments under a post
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 2816 entries, 0 to 2815
Data columns (total 11 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 image_ids 2816 non-null object
1 comment_paths 2816 non-null object
2 id 2814 non-null float64
3 created_at 2814 non-null float64
4 text 2814 non-null object
5 likes_count 2814 non-null float64
6 answers 2814 non-null object
7 id.1 2814 non-null float64 # ID of the user who commented
8 is_verified 2814 non-null object
9 profile_pic_url 2814 non-null object
10 username 2814 non-null object
dtypes: float64(4), object(7)
memory usage: 242.1+ KB
posts_text.csv - Contains all the posts with their text extracted through their image using OCR(Optical Character Recognition)
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 159 entries, 0 to 158
Data columns (total 7 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 hashtags 159 non-null object
1 captions 139 non-null object
2 likes 159 non-null int64
3 comments 159 non-null int64
4 image_ids 159 non-null object
5 image_paths 159 non-null object
6 image_text 159 non-null object
dtypes: int64(2), object(5)
memory usage: 8.8+ KB
This is an issue with your PYTHONPATH, setting it to something like export PYTHONPATH="${PYTHONPATH}:/Users/yravindranath/REPO" should fix it.
- So currently the entire project isn't repoducible therefore I will dockerise it soon and allow anyone to run it locally without any issues.
- If you notice there is a slow
apply()used for binarizing the images and extracting the text from it using OCR. I am usingswifterto speed it up as it is.