Implementation for the paper submitted to The 5th IEEE International Conference on Multimedia Information Processing and Retrieval (MIPR 2022) MCAUIS Workshop.
Automatic Evaluation of Machine Generated Feedback For Text and Image Data
Pratham Goyal*, Anjali Raj*, Puneet Kumar, and Balasubramanian Raman
- Install Anaconda or Miniconda distribution and create a conda environment with Python 3.6+.
- Install the requirements using the following command:
pip install -r Requirements.txt- Download glove.6B.zip, unzip and keep in
glove.6Bfolder. - Download the required datasets.
- We split the datatset into three parts and apply pre-processing techniques to clean the data.We have three sets of csv files comprising of training data, validation data and test data.
- The test feedbacks are generated by running the pre-trained model described in the [Base Paper](add link). We get the csv file
{{Epoch_Number}}Ep_test_results.csv{{Epoch_Number}}Ep_test_feedbacks.csvand{{Epoch_Number}}Ep_test_comments.csvas a result.These feedbacks are then fed into the Similarity Module to generate the similarity scores.
BERT Steps to run the Code
...
Access to the ‘IIT Roorkee Multimodal Feedback Synthesis (IIT-R MMFeed) dataset’ can be obtained by through Access Form - IIT-R MMFeed Dataset.pdf. The dataset is prepared at Machine Intelligence Lab, IIT Roorkee under the supervision of Prof. Balasubramanian Raman. It consists of 9,479 samples containing news text, images, user comments, and the number of likes for each comment.