update file structures

README.md

@@ -2,6 +2,31 @@
 license: mit
 ---
 
+---
+configs:
+- config_name: ordered_id_splits
+  data_files:
+  - split: train
+    path:
+    - "ordered_id_splits/valid_input.tsv"
+    - "ordered_id_splits/valid_target.tsv"
+    - "ordered_id_splits/test_input.tsv"
+  - split: test
+    path: "ordered_id_splits/ghi.csv"
+  default: true
+- config_name: interaction_splits
+  data_files:
+  - split: valid
+    path:
+    - "interaction_splits/valid_inter_input.tsv"
+    - "interaction_splits/valid_inter_target.csv"
+  - split: test
+    path: "interaction_splits/test_input.tsv"
+- config_name: clueweb_id_mapping
+  data_files: "cwid_to_id.tsv"
+---
+
+
 
 # ClueWeb-Reco: 
 

cw_data_processing/ClueWeb22Api.py

@@ -0,0 +1,143 @@
+import os
+import gzip
+import zipfile
+
+class ClueWeb22Api:
+
+    """
+    ClueWeb22Api adapted from source: https://github.com/lemurproject/ClueWeb22/blob/main/ClueWeb22Api.py
+    """
+
+    def __init__(self, cw22id, cw22root_path):
+        self.cw22id = cw22id
+        self.cw22root_path = cw22root_path
+
+    def get_base_filename_by_id(self, cw22id, cw22root_path, file_type='html'):
+        # /data/datasets/clueweb22/ClueWeb22_B/txt
+        html_path = self.cw22root_path + os.sep + file_type
+
+        # clueweb22-en00{subfolder_id}-{jjsongz_id}-{ddoc_id}"
+        id_parts = cw22id.split('-')
+        doc = int(id_parts[len(id_parts) - 1])
+
+        language = id_parts[1][:2] # en 
+        segment = id_parts[1][:4] # en00 
+        directory = id_parts[1] # en0000
+        # /data/datasets/clueweb22/ClueWeb22_B/txt + en + en00 + en0000 
+        base_path = html_path + os.sep + language + os.sep + segment + os.sep + directory + os.sep
+        base_filename = base_path + id_parts[1] + '-' + id_parts[2]
+        return base_filename
+
+    def get_primary_node_ids(self, annotate_html):
+        annotations = annotate_html.annotations
+        primary_node_ids = []
+        for annotation in annotations:
+            if annotation.type == AnnotateHtml.AnnotationType.Primary:
+                primary_node_ids.append(int(annotation.nodeId))
+        primary_node_ids.sort()
+        return primary_node_ids
+
+    def get_html_from_warc(self):
+        cw22id = self.cw22id
+        cw22root_path = self.cw22root_path
+        base_filename = self.get_base_filename_by_id(cw22id, cw22root_path)
+
+        warc_path = base_filename + '.warc.gz'
+        offset_path = base_filename + '.warc.offset'
+
+        id_parts = cw22id.split('-')
+        doc = int(id_parts[len(id_parts) - 1])
+
+        #Get html from warc using offset
+        offset_length = len('{:010d}\n'.format(0, 0))
+        with open (warc_path,'rb') as f_warc:
+            with open (offset_path, 'r') as f_offset:
+                f_offset.seek(int(doc) * int(offset_length))
+                start_bytes = int (f_offset.read (offset_length).strip())
+                end_bytes =   int (f_offset.read (offset_length).strip())
+                f_warc.seek(start_bytes)
+                record = f_warc.read(end_bytes - start_bytes)
+                record = gzip.decompress(record).decode('utf-8')
+
+                #Remove the WARC header to get the htmlStr
+                warc_header = ''
+                for line in record.splitlines():
+                    warc_header += line
+                    warc_header += '\r\n'
+                    if len(line.strip()) == 0:
+                        break
+                record = record[len(warc_header):]
+
+                return record
+    
+
+
+    def get_json_record(self, record_type):
+        cw22id = self.cw22id
+        cw22root_path = self.cw22root_path
+        base_filename = self.get_base_filename_by_id(cw22id, cw22root_path, file_type=record_type)
+        json_path = base_filename + '.json.gz'
+        offset_path = base_filename + '.offset'
+
+        id_parts = cw22id.split('-')
+        doc = int(id_parts[len(id_parts) - 1])
+
+        offset_length = len('{:010d}\n'.format(0, 0))
+        with open (json_path,'rb') as f_json:
+            with open (offset_path, 'r') as f_offset:
+                f_offset.seek(int(doc) * int(offset_length))
+                start_bytes = int (f_offset.read (offset_length).strip())
+                end_bytes =   int (f_offset.read (offset_length).strip())
+                f_json.seek(start_bytes)
+                record = f_json.read(end_bytes - start_bytes)
+                record = gzip.decompress(record).decode('utf-8')
+                return record
+
+
+    def get_clean_text(self):
+        record = self.get_json_record('txt')
+        return record
+
+    def get_inlinks(self):
+        record = self.get_json_record('inlink')
+        return record
+
+    def get_outlinks(self):
+        record = self.get_json_record('outlink')
+        return record
+
+
+    def get_marcoweb_clean_text(self):
+            record = self.get_marcoweb_json_record('txt')
+            return record
+        
+    def get_marcoweb_json_record(self, record_type):
+        cw22id = self.cw22id
+        cw22root_path = self.cw22root_path
+
+        def get_base_filename():
+            html_path = cw22root_path + os.sep 
+            id_parts = cw22id.split('-') # "clueweb22-en-{jjsongz_id}-{ddoc_id}"
+            language = id_parts[1]
+            base_path = html_path + os.sep + language + os.sep 
+            base_filename = base_path + id_parts[1] + '-' + id_parts[2]
+            return base_filename
+
+        # custom file path 
+        base_filename = get_base_filename()
+        json_path = base_filename + '.json.gz'
+        offset_path = base_filename + '.offset'
+
+        id_parts = cw22id.split('-')
+        doc = int(id_parts[len(id_parts) - 1])
+
+        offset_length = len('{:010d}\n'.format(0, 0))
+        with open (json_path,'rb') as f_json:
+            with open (offset_path, 'r') as f_offset:
+                f_offset.seek(int(doc) * int(offset_length))
+                start_bytes = int (f_offset.read (offset_length).strip())
+                end_bytes =   int (f_offset.read (offset_length).strip())
+                f_json.seek(start_bytes)
+                record = f_json.read(end_bytes - start_bytes)
+                record = gzip.decompress(record).decode('utf-8')
+                return record

cw_data_processing/example_dataset.py

@@ -0,0 +1,56 @@
+from ClueWeb22Api import ClueWeb22Api
+from tqdm import tqdm 
+
+
+def get_clueweb_title(cwid, clueweb_path): 
+    clueweb_api = ClueWeb22Api(cwid, clueweb_path)
+    clean_txt = eval(clueweb_api.get_clean_text())
+    content = clean_txt["Clean-Text"]
+    title = content.split('\n')[0].replace("\n", "").replace("\t", "").replace("\r", "").replace("\'", "").replace("\"", "").strip()
+    return title    
+
+    
+def load_data_clueweb(filename, id_map_path, clueweb_path): 
+    """
+        args: 
+            filename: the validation set input (valid.tsv) or the test set input (test.tsv)
+            id_map_path: the mapping from official clueweb id to our internal clueweb id (cwid_to_id.tsv) 
+            clueweb_path: your clueweb dataset path on disk (ex. /../data/datasets/clueweb22/ClueWeb22_B)
+    """
+
+    # load mapping to official clueweb doc ids 
+    id_to_cwid = {}
+    with open(id_map_path, "r") as f:
+        for line in f:  
+            parts = line.strip().split("\t")
+            id_to_cwid[int(parts[1])] = parts[0]
+
+
+    data = {}
+    lines = open(filename, 'r').readlines()
+    for line in tqdm(lines[1:]):
+        history_titles = list()
+        line = line.strip().split('\t')
+        
+        session_id = line[0]
+        history = line[1].split(",")
+
+        for internal_id in history: 
+            internal_id = int(internal_id)
+            # get the title of the item by ClueWebApi 
+            title = get_clueweb_title(id_to_cwid[internal_id], clueweb_path)
+            history_titles.append(title)
+
+        data[session_id] = history_titles
+
+    return data
+
+
+id_map_path="ClueWeb-Reco/cwid_to_id.tsv"
+seq_data_path="ClueWeb-Reco_public/ordered_id_splits/valid_input.tsv"
+clueweb_path="/data/datasets/clueweb22/ClueWeb22_B"
+
+# dictionary with session_ids as key, list of strings of the titles of historically interacted item
+data = load_data_clueweb(filename=seq_data_path, id_map_path=id_map_path, clueweb_path=clueweb_path)
+
+