data_process/data_construction.py

import json
import argparse

def parse_args():
    parser = argparse.ArgumentParser()
    parser.add_argument('--review_data_path', type=str, default='./data')
    parser.add_argument('--meta_data_path', type=str, default='./data')
    parser.add_argument('--save_dir', type=str, default='./data/dataset')
    parser.add_argument('--dataset', type=str, default='dataset')
    return parser.parse_args()

args = parse_args()

import os
if not os.path.exists(args.save_dir + '/' + args.dataset):
    os.mkdir(args.save_dir + '/' + args.dataset)
# /datain/v-yinju/LLMBased_Multimodal_RS/Data/Musical_Instruments

''' Extract interaction sequence '''
inters = {}
with open(args.review_data_path, 'r', encoding = 'utf-8') as file:
# with open('/datain/v-yinju/LLMBased_Multimodal_RS/Data/Musical_Instruments.json', 'r', encoding = 'utf-8') as file:
    for line in file:
        element = json.loads(line)
        if element['reviewerID'] in inters:
            inters[element['reviewerID']].append({'time': element['unixReviewTime'], 'item': element['asin']})
        else:
            inters[element['reviewerID']] = [{'time': element['unixReviewTime'], 'item': element['asin']}]
# Filter out sequence shorter than 5
filtered_inters = {key: value for key, value in inters.items() if len(value) > 4}
final_inters = {}
for key, value in filtered_inters.items():
    # Sort items according to time
    value.sort(key = lambda x: x['time'])
    final_inters[key] = [x['item'] for x in value]
# Save interaction
with open(args.save_dir + '/' + args.dataset + '/' + args.dataset + '.inters.json', 'w', encoding = 'utf-8') as f:
    json.dump(final_inters, f, ensure_ascii = False, indent = 4)

''' Extract user review '''
reviews = {}
with open(args.review_data_path, 'r', encoding = 'utf-8') as file:
# with open('/datain/v-yinju/LLMBased_Multimodal_RS/Data/Musical_Instruments.json', 'r', encoding = 'utf-8') as file:
    for line in file:
        element = json.loads(line)
        if len(element.get('reviewText', '')) > 0:
            reviews[element['reviewerID']] = {element['asin']: element['reviewText']}
        else:
            continue
with open(args.save_dir + '/' + args.dataset + '/' + args.dataset + '.reviews.json', 'w', encoding = 'utf-8') as f:
    json.dump(reviews, f, ensure_ascii = False, indent = 4)

''' Extract item features '''
features = {}
with open(args.meta_data_path, 'r', encoding = 'utf-8') as file:
# with open('/datain/v-yinju/LLMBased_Multimodal_RS/Data/meta_Musical_Instruments.json', 'r', encoding = 'utf-8') as file:
    for line in file:
        element = json.loads(line)
        if len(element.get('title', '')) > 0 and len(element.get('description', '')) > 0 and len(element.get('imageURL', '')) > 0 and len(element.get('imageURLHighRes', '')) > 0:
            features[element['asin']] = {
                'title': element['title'], 
                'description': element['description'], 
                'image': element['imageURL'], 
                'imageH': element['imageURLHighRes']}
        else:
            continue
with open(args.save_dir + '/' + args.dataset + '/' + args.dataset + '.features.json', 'w', encoding = 'utf-8') as f:
    json.dump(features, f, ensure_ascii = False, indent = 4)