Statistics is missing multi version of moviegraphbenchmark (#40)

* Add MGB multi setting to statistics creation

* Move statistics function, adapt and update csv

* Adapt import and README

* Move import again

README.md

@@ -158,7 +158,7 @@ Datasets
 More broad statistics are provided in `dataset_statistics.csv`. You can also get a pandas DataFrame with statistics for specific datasets for example to create tables for publications:
 ```
 >>> ds = MovieGraphBenchmark(graph_pair="multi")
->>> from sylloge.base import create_statistics_df
+>>> from sylloge.create_statistic import create_statistics_df
 >>> stats_df = create_statistics_df([ds])
 >>> stats_df.loc[("MovieGraphBenchmark","moviegraphbenchmark_multi","imdb")]
                                                             Entities  Relation Triples  Attribute Triples  ...  Clusters  Intra-dataset Matches  All Matches

dataset_statistics.csv

@@ -37,8 +37,11 @@ MovieGraphBenchmark,moviegraphbenchmark_imdb_tvdb,imdb,5129,17507,20800,3,13,608
 MovieGraphBenchmark,moviegraphbenchmark_imdb_tvdb,tvdb,7814,15455,20902,3,9,7683,1483,22663,25583
 MovieGraphBenchmark,moviegraphbenchmark_tmdb_tvdb,tmdb,6061,27903,23761,4,30,9991,1920,64,26138
 MovieGraphBenchmark,moviegraphbenchmark_tmdb_tvdb,tvdb,7814,15455,20902,3,9,7683,1920,22663,26138
-MED_BBK,med_bbk,MED,9162,158357,11467,32,19,10858,8885,0,5619
-MED_BBK,med_bbk,BBK,9162,50307,44987,20,21,36608,8885,0,5619
+MovieGraphBenchmark,moviegraphbenchmark_multi,imdb,5129,17507,20800,3,13,6082,3598,1,31230
+MovieGraphBenchmark,moviegraphbenchmark_multi,tmdb,6061,27903,23761,4,30,9991,3598,64,31230
+MovieGraphBenchmark,moviegraphbenchmark_multi,tvdb,7814,15455,20902,3,9,7683,3598,22663,31230
+MED_BBK,med_bbk,MED,9162,158357,11467,32,19,10858,9162,0,9162
+MED_BBK,med_bbk,BBK,9162,50307,44987,20,21,36608,9162,0,9162
 OAEI,oaei_marvelcinematicuniverse_marvel,marvelcinematicuniverse,216033,1094598,130517,130,110,56566,1654,0,1654
 OAEI,oaei_marvelcinematicuniverse_marvel,marvel,1472619,5152898,1580468,63,127,749980,1654,0,1654
 OAEI,oaei_memoryalpha_memorybeta,memoryalpha,254537,2096198,430730,180,287,226110,9296,0,9296

sylloge/__init__.py

@@ -10,7 +10,6 @@
     TrainTestValSplit,
     ZipEADataset,
     ZipEADatasetWithPreSplitFolds,
-    create_statistics_df,
 )
 from .id_mapped import IdMappedEADataset
 from .med_bbk_loader import MED_BBK
@@ -35,7 +34,6 @@
     "BinaryZipEADatasetWithPreSplitFolds",
     "ZipEADatasetWithPreSplitFolds",
     "TrainTestValSplit",
-    "create_statistics_df",
 ]
 __version__ = version(__package__)
 logging.getLogger(__name__).setLevel(logging.INFO)

sylloge/base.py

@@ -11,7 +11,6 @@
     Callable,
     Dict,
     Generic,
-    Iterable,
     List,
     Literal,
     Mapping,
@@ -1085,58 +1084,3 @@ class BinaryZipEADatasetWithPreSplitFolds(
 
     def __repr__(self) -> str:
         return self._binary_repr_adjustment(super().__repr__())
-
-
-def create_statistics_df(
-    datasets: Iterable[MultiSourceEADataset], seperate_attribute_relations: bool = True
-):
-    rows = []
-    triples_col = (
-        ["Relation Triples", "Attribute Triples"]
-        if seperate_attribute_relations
-        else ["Triples"]
-    )
-    index_cols = ["Dataset family", "Task Name", "Dataset Name"]
-    columns = [
-        *index_cols,
-        "Entities",
-        *triples_col,
-        "Relations",
-        "Properties",
-        "Literals",
-        "Clusters",
-        "Intra-dataset Matches",
-        "All Matches",
-    ]
-    for ds in datasets:
-        ds_family = str(ds.__class__.__name__).split(".")[-1]
-        ds_stats, num_clusters = ds.statistics()
-        all_matches = ds.ent_links.number_of_links
-        intra_dataset_matches = (0,) * len(ds.dataset_names)
-        if isinstance(ds.ent_links, PrefixedClusterHelper):
-            intra_dataset_matches = ds.ent_links.number_of_intra_links
-        for i, (ds_side, ds_side_name) in enumerate(zip(ds_stats, ds.dataset_names)):
-            if seperate_attribute_relations:
-                triples = [ds_side.rel_triples, ds_side.attr_triples]
-            else:
-                triples = [ds_side.triples]
-            rows.append(
-                [
-                    ds_family,
-                    ds.canonical_name,
-                    ds_side_name,
-                    ds_side.entities,
-                    *triples,
-                    ds_side.relations,
-                    ds_side.properties,
-                    ds_side.literals,
-                    num_clusters,
-                    intra_dataset_matches[i],
-                    all_matches,
-                ]
-            )
-    statistics_df = pd.DataFrame(
-        rows,
-        columns=columns,
-    )
-    return statistics_df.set_index(index_cols)

sylloge/create_statistic.py

@@ -1,9 +1,68 @@
 from typing import Dict, Iterable, Tuple
 
 import pandas as pd
+from eche import ClusterHelper, PrefixedClusterHelper
 
 from sylloge import MED_BBK, OAEI, MovieGraphBenchmark, MultiSourceEADataset, OpenEA
-from sylloge.base import create_statistics_df
+
+
+def create_statistics_df(
+    datasets: Iterable[MultiSourceEADataset], seperate_attribute_relations: bool = True
+):
+    rows = []
+    triples_col = (
+        ["Relation Triples", "Attribute Triples"]
+        if seperate_attribute_relations
+        else ["Triples"]
+    )
+    index_cols = ["Dataset family", "Task Name", "Dataset Name"]
+    columns = [
+        *index_cols,
+        "Entities",
+        *triples_col,
+        "Relations",
+        "Properties",
+        "Literals",
+        "Clusters",
+        "Intra-dataset Matches",
+        "All Matches",
+    ]
+    for ds in datasets:
+        ds_family = str(ds.__class__.__name__).split(".")[-1]
+        ds_stats, num_clusters = ds.statistics()
+        intra_dataset_matches = (0,) * len(ds.dataset_names)
+        if isinstance(ds.ent_links, ClusterHelper):
+            all_matches = ds.ent_links.number_of_links
+            if isinstance(ds.ent_links, PrefixedClusterHelper):
+                intra_dataset_matches = ds.ent_links.number_of_intra_links
+        else:
+            all_matches = len(ds.ent_links)
+        for i, (ds_side, ds_side_name) in enumerate(zip(ds_stats, ds.dataset_names)):
+            if seperate_attribute_relations:
+                triples = [ds_side.rel_triples, ds_side.attr_triples]
+            else:
+                triples = [ds_side.triples]
+            rows.append(
+                [
+                    ds_family,
+                    ds.canonical_name,
+                    ds_side_name,
+                    ds_side.entities,
+                    *triples,
+                    ds_side.relations,
+                    ds_side.properties,
+                    ds_side.literals,
+                    num_clusters,
+                    intra_dataset_matches[i],
+                    all_matches,
+                ]
+            )
+    statistics_df = pd.DataFrame(
+        rows,
+        columns=columns,
+    )
+    return statistics_df.set_index(index_cols)
+
 
 all_classes_with_args: Tuple[Tuple[type[MultiSourceEADataset], Dict[str, str]], ...] = (
     (OpenEA, {"graph_pair": "D_W", "size": "15K", "version": "V1"}),
@@ -25,6 +84,7 @@
     (MovieGraphBenchmark, {"graph_pair": "imdb-tmdb"}),
     (MovieGraphBenchmark, {"graph_pair": "imdb-tvdb"}),
     (MovieGraphBenchmark, {"graph_pair": "tmdb-tvdb"}),
+    (MovieGraphBenchmark, {"graph_pair": "multi"}),
     (MED_BBK, {}),
     (OAEI, {"task": "marvelcinematicuniverse-marvel"}),
     (OAEI, {"task": "memoryalpha-memorybeta"}),
@@ -34,7 +94,7 @@
 )
 
 
-def create_statistic(
+def create_and_write_statistic(
     classes_with_args: Iterable[
         Tuple[type[MultiSourceEADataset], Dict[str, str]]
     ] = all_classes_with_args,
@@ -49,4 +109,4 @@ def create_statistic(
 
 
 if __name__ == "__main__":
-    create_statistic()
+    create_and_write_statistic()