Upload 2 files

Twitter/processed/Twitter-processing.py

@@ -0,0 +1,152 @@
+import pandas as pd
+import numpy as np
+import pickle as pkl
+import networkx as nx
+
+from torch_geometric.data import Data
+import torch
+
+# Define file paths
+p_part1 = 'processed/original/68841_tweets_multiclasses_filtered_0722_part1.npy'
+p_part2 = 'processed/original/68841_tweets_multiclasses_filtered_0722_part2.npy'
+
+# Load the numpy data
+df_np_part1 = np.load(p_part1, allow_pickle=True)
+df_np_part2 = np.load(p_part2, allow_pickle=True)
+df_np = np.concatenate((df_np_part1, df_np_part2), axis=0)
+
+# Convert numpy data to pandas DataFrame
+df = pd.DataFrame(data=df_np, columns=["event_id", "tweet_id", "text", "user_id", "created_at", "user_loc",
+    "place_type", "place_full_name", "place_country_code", "hashtags", "user_mentions", "image_urls", "entities",
+    "words", "filtered_words", "sampled_words"])
+print("Data converted to dataframe.")
+
+# Select relevant columns
+df_graph_4 = df[['tweet_id', 'text', 'user_id', 'user_mentions', 'event_id']]
+
+def create_graph_4(df_graph):
+    # Create an empty graph
+    G = nx.Graph()
+    text_nodes = []
+    edge_index = [[], []]
+    text_edges = []
+    node_labels = []
+    edge_labels = []
+
+    # Create all user nodes and user_id2idx
+    user_id2idx = {}
+    user_nodes = []
+    for _, row in df_graph.iterrows():
+        mentions_list = row['user_mentions']
+        user_id = row["user_id"]
+        # Only include user nodes where users mention each other
+        if len(mentions_list) > 0:
+            if user_id not in user_id2idx:
+                user_id2idx[user_id] = len(user_nodes)
+                user_nodes.append({user_id: user_id2idx[user_id]})
+                G.add_node(user_id2idx[user_id])
+            for mention in mentions_list:
+                if mention not in user_id2idx:
+                    user_id2idx[mention] = len(user_nodes)
+                    user_nodes.append({mention: user_id2idx[mention]})
+                    G.add_node(user_id2idx[mention])
+
+    text_nodes = ["user"] * len(user_nodes)  # All user nodes have text "user" and label -1
+    node_labels = [-1] * len(user_nodes)
+    print("Length of user nodes:", len(user_nodes))
+    print("Sample user nodes:", user_nodes[:5])
+    print("Sample user node labels:", node_labels[:5])
+    print("Sample user node texts:", text_nodes[:5])
+
+    # Initialize tweet_id2idx
+    tweet_id2idx = {}
+    tweet_id2node_idx = {}
+    tweet_nodes = []
+
+    # Add information related to mentions (user-user edge, tweet node, user-tweet edge)
+    for _, row in df_graph.iterrows():
+        mentions_list = row['user_mentions']
+        if len(mentions_list) > 0:
+            user_idx = user_id2idx[row['user_id']]
+
+            for mention in mentions_list:
+                mention_idx = user_id2idx[mention]
+                if not G.has_edge(user_idx, mention_idx):  # Only include one edge between two users
+                    # Add edge u1-u2
+                    G.add_edge(user_idx, mention_idx)
+                    edge_index[0].append(user_idx)
+                    edge_index[1].append(mention_idx)
+                    text_edges.append(row['text'])
+                    edge_labels.append(row["event_id"])
+
+                    # Add node t1, create tweet_id2idx
+                    tweet_id = row['tweet_id']
+                    tweet_id2idx[tweet_id] = len(tweet_nodes)
+                    tweet_id2node_idx[tweet_id] = len(user_nodes) + tweet_id2idx[tweet_id]
+                    tweet_nodes.append({tweet_id: len(user_nodes) + tweet_id2idx[tweet_id]})
+                    G.add_node(tweet_id2node_idx[tweet_id])
+                    text_nodes.append(row["text"])
+                    node_labels.append(row['event_id'])
+
+                    # Add edge u1-t1
+                    tweet_node_idx = tweet_id2node_idx[tweet_id]
+                    edge_index[0].append(user_idx)
+                    edge_index[1].append(tweet_node_idx)
+                    text_edges.append("")
+                    edge_labels.append(-1)
+                    G.add_edge(user_idx, tweet_node_idx)
+
+                    # Add edge t1-u2
+                    edge_index[0].append(mention_idx)
+                    edge_index[1].append(tweet_node_idx)
+                    text_edges.append("")
+                    edge_labels.append(-1)
+                    G.add_edge(mention_idx, tweet_node_idx)
+
+    # Add information not related to mentions (user-tweet edge, tweet node)
+    for _, row in df_graph.iterrows():
+        tweet_id = row['tweet_id']
+        user_id = row["user_id"]
+        if user_id in user_id2idx.keys() and tweet_id not in tweet_id2idx.keys():
+            # Create tweet_id2node_idx and add node t1
+            tweet_id2idx[tweet_id] = len(tweet_nodes)
+            tweet_id2node_idx[tweet_id] = len(user_nodes) + tweet_id2idx[tweet_id]
+            tweet_nodes.append({tweet_id: len(user_nodes) + tweet_id2idx[tweet_id]})
+            G.add_node(tweet_id2node_idx[tweet_id])
+            text_nodes.append(row["text"])
+            node_labels.append(row['event_id'])
+
+            # Add edge u1-t1
+            user_idx = user_id2idx[user_id]
+            tweet_node_idx = tweet_id2node_idx[tweet_id]
+            edge_index[0].append(user_idx)
+            edge_index[1].append(tweet_node_idx)
+            text_edges.append("")
+            edge_labels.append(-1)
+            G.add_edge(user_idx, tweet_node_idx)
+
+    print("Length of tweet nodes:", len(tweet_nodes))
+    print("Sample tweet nodes:", tweet_nodes[:5])
+    print("Sample tweet node labels:", node_labels[-5:])
+    print("Sample tweet node texts:", text_nodes[-5:])
+
+    return text_nodes, text_edges, node_labels, edge_labels, edge_index
+
+# Create the graph
+text_nodes, text_edges, node_labels, edge_labels, edge_index = create_graph_4(df_graph_4)
+
+# Create Data object
+graph = Data(
+    text_nodes=text_nodes,
+    text_edges=text_edges,
+    node_labels=torch.tensor(node_labels, dtype=torch.long),
+    edge_labels=torch.tensor(edge_labels, dtype=torch.long),
+    edge_index=torch.tensor(edge_index, dtype=torch.long)
+)
+
+# Save the processed graph data
+output_file = os.path.join(base_dir, 'output/twitter_graph.pkl')
+with open(output_file, 'wb') as file:
+    pkl.dump(graph, file)
+
+print(f"Data processing complete. Processed data saved to: {output_file}")

Twitter/processed/download_data.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+
+# Get the directory of the current script
+SCRIPT_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )"
+BASE_DIR="$(dirname "$SCRIPT_DIR")"
+RAW_DIR="$BASE_DIR/processed/original"
+
+# Create the raw directory
+mkdir -p "$RAW_DIR"
+
+# Define URLs of the files to be downloaded
+urls=(
+    "https://github.com/YuweiCao-UIC/KPGNN/raw/main/datasets/Twitter/68841_tweets_multiclasses_filtered_0722_part1.npy"
+    "https://github.com/YuweiCao-UIC/KPGNN/raw/main/datasets/Twitter/68841_tweets_multiclasses_filtered_0722_part2.npy"
+)
+
+# Download each file to the raw directory
+for url in "${urls[@]}"; do
+    wget -P "$RAW_DIR" "$url"
+done
+
+echo "Download complete."
+
+
+