MDAT
A multimodal data annotation tool: For annotation of multimodal (text and image) Twitter posts.
How to use
Using MDAT is simple, make sure you are connected to the internet. Follow the steps below:
Step 1: Go to the link (https://drive.google.com/file/d/1eG5sECQDdHkVTS8P72M-PB-hWgb29iLr/view?usp=sharing) and download the MDAT.rar file
Step 2: Locate and extract the file somewhere in PC.
Step 3: Run the exe file (will take some time first time because it will be downloading some important modules)
Step 4: Upload the csv/xlsx file containing Tweets (Tweet_ID, Tweet_Text, Media_URL) (example file is given in data.xlsx)
Step 5: Annotate the data with the following process:
Text annotation (semi-automatic):
a. Choose, if the post is relevant or irrelevant
b. If relevant, look at the automatic target, sentiment suggestions provided by system.
If annotations are correct move to next step, otherwise correct the target or sentiment.
Image annotation (manual):
c. Select the sentiment region(s) in the image. (Use previous or next button if a wrong region is selected)
d. Enter if there is any target in the image, and select sentiment towards the target.
d. Check for the overall sentiment of the post, correct it if needed.
e. Save the annotation for each post. (Saving is important)
To move previous or next tweet, make use of previous and next button on the screen.
Enjoy creating your own multimodal dataset.