Natural language processing course: Automatic generation of Slovenian traffic news for RTV Slovenija

1. Requirements

• folder inside project root named outputs with all wanted rtf report files
• csv file named podatki inside project code folder with csv data for the same dates as the rtf reports

2. Local data setup

When all of the requirements are setup you will have to run code/test_distance.py. This will generate a file named useful_matches.json.

3. Running on HPC

We have already done steps 1 and 2. So you only need to do this one.

Connect to HPC and move to /d/hpc/projects/onj_fri/vmzteam. Then run run_slurm.sh.

This will create a file called similarity_scores.csv which contains the calculated BLUE and BERT scores of the generated reports. And it will also create a folder named llm_outputs in the vmzteam folder that contains the generated reports.