Update etpc.py

etpc.py

@@ -16,13 +16,12 @@
 """ETPC: The Extended Typology Paraphrase Corpus"""
 
 import os
-from typing import Dict, Generator, Any, Union, Optional, List, Tuple
+from typing import Any, Dict, Generator, List, Optional, Tuple, Union
 
 import datasets
 from datasets.tasks import TextClassification
 from lxml import etree
 
-
 logger = datasets.logging.get_logger(__name__)
 
 
@@ -63,13 +62,22 @@ class ETPC(datasets.GeneratorBasedBuilder):
     def _info(self):
         features = datasets.Features(
             {
-                "idx": datasets.Value("int32"),
+                "idx": datasets.Value("string"),
                 "sentence1": datasets.Value("string"),
                 "sentence2": datasets.Value("string"),
                 "etpc_label": datasets.Value("int8"),
                 "mrpc_label": datasets.Value("int8"),
                 "negation": datasets.Value("int8"),
-                "paraphrase_types": datasets.Value("string"),
+                "paraphrase_type": datasets.Value("string"),
+                "paraphrase_type_id": datasets.Value("int32"),
+                "sentence1_phrase_span": datasets.Sequence(
+                    datasets.Value("int32")
+                ),
+                "sentence2_phrase_span": datasets.Sequence(
+                    datasets.Value("int32")
+                ),
+                "sentence1_phrase_text": datasets.Value("string"),
+                "sentence2_phrase_text": datasets.Value("string"),
             }
         )
 
@@ -101,26 +109,66 @@ class ETPC(datasets.GeneratorBasedBuilder):
         parser = etree.XMLParser(encoding="utf-8", recover=True)
 
         tree_text_pairs = etree.parse(text_pairs_path, parser=parser)
-        tree_paraphrase_types = etree.parse(paraphrase_types_path, parser=parser)
+        tree_paraphrase_types = etree.parse(
+            paraphrase_types_path, parser=parser
+        )
 
         root_text_pairs = tree_text_pairs.getroot()
         root_paraphrase_types = tree_paraphrase_types.getroot()
 
-        for idx, row in enumerate(root_text_pairs):
-            children = row.getchildren()
+        idx = -1
+
+        for row in root_text_pairs:
             current_pair_id = row.find(".//pair_id").text
             paraphrase_types = root_paraphrase_types.xpath(
                 f".//pair_id[text()='{current_pair_id}']/parent::relation/type_name/text()"
             )
-            str_paraphrase_types = ",".join(paraphrase_types)
-            yield idx, {
-                "idx": int(row.find(".//pair_id").text)
-                if row.find(".//pair_id") is not None
-                else int(idx),
-                "sentence1": row.find(".//sent1_raw").text,
-                "sentence2": row.find(".//sent2_raw").text,
-                "etpc_label": int(row.find(".//etpc_label").text),
-                "mrpc_label": int(row.find(".//mrpc_label").text),
-                "negation": int(row.find(".//negation").text),
-                "paraphrase_types": str_paraphrase_types,
-            }
+            paraphrase_type_ids = root_paraphrase_types.xpath(
+                f".//pair_id[text()='{current_pair_id}']/parent::relation/type_id/text()"
+            )
+            sentence1_phrase_text = root_paraphrase_types.xpath(
+                f".//pair_id[text()='{current_pair_id}']/parent::relation/s1_text/text()"
+            )
+            sentence2_phrase_text = root_paraphrase_types.xpath(
+                f".//pair_id[text()='{current_pair_id}']/parent::relation/s2_text/text()"
+            )
+            sentence1_phrase_spans = root_paraphrase_types.xpath(
+                f".//pair_id[text()='{current_pair_id}']/parent::relation/s1_scope/text()"
+            )
+            sentence2_phrase_spans = root_paraphrase_types.xpath(
+                f".//pair_id[text()='{current_pair_id}']/parent::relation/s2_scope/text()"
+            )
+            for (
+                paraphrase_type,
+                paraphrase_type_id,
+                sentence1_phrase_span,
+                sentence2_phrase_span,
+                sentence1_phrase_text,
+                sentence2_phrase_text,
+            ) in zip(
+                paraphrase_types,
+                paraphrase_type_ids,
+                sentence1_phrase_spans,
+                sentence2_phrase_spans,
+                sentence1_phrase_text,
+                sentence2_phrase_text,
+            ):
+                idx += 1
+                yield idx, {
+                    "idx": row.find(".//pair_id").text
+                    + "_"
+                    + paraphrase_type_id
+                    + "_"
+                    + str(idx),
+                    "sentence1": row.find(".//sent1_raw").text,
+                    "sentence2": row.find(".//sent2_raw").text,
+                    "etpc_label": int(row.find(".//etpc_label").text),
+                    "mrpc_label": int(row.find(".//mrpc_label").text),
+                    "negation": int(row.find(".//negation").text),
+                    "paraphrase_type": paraphrase_type,
+                    "paraphrase_type_id": paraphrase_type_id,
+                    "sentence1_phrase_span": sentence1_phrase_span.split(","),
+                    "sentence2_phrase_span": sentence2_phrase_span.split(","),
+                    "sentence1_phrase_text": sentence1_phrase_text,
+                    "sentence2_phrase_text": sentence2_phrase_text,
+                }