fix readme

.gitattributes

@@ -80,3 +80,19 @@ data/t_rex.filter.min_entity_16_max_predicate_100.jsonl filter=lfs diff=lfs merg
 data/t_rex.filter.min_entity_4_max_predicate_25.jsonl filter=lfs diff=lfs merge=lfs -text
 data/t_rex.filter.min_entity_8_max_predicate_25.jsonl filter=lfs diff=lfs merge=lfs -text
 data/t_rex.filter_unified.jsonl filter=lfs diff=lfs merge=lfs -text
+data/t_rex.filter_unified.min_entity_16_max_predicate_25.jsonl filter=lfs diff=lfs merge=lfs -text
+data/t_rex.filter_unified.min_entity_8_max_predicate_100.jsonl filter=lfs diff=lfs merge=lfs -text
+data/t_rex.filter_unified.min_entity_12_max_predicate_25.jsonl filter=lfs diff=lfs merge=lfs -text
+data/t_rex.filter_unified.min_entity_8_max_predicate_10.jsonl filter=lfs diff=lfs merge=lfs -text
+data/t_rex.filter_unified.min_entity_8_max_predicate_25.jsonl filter=lfs diff=lfs merge=lfs -text
+data/t_rex.filter_unified.min_entity_8_max_predicate_50.jsonl filter=lfs diff=lfs merge=lfs -text
+data/t_rex.filter_unified.min_entity_4_max_predicate_50.jsonl filter=lfs diff=lfs merge=lfs -text
+data/t_rex.filter_unified.min_entity_16_max_predicate_10.jsonl filter=lfs diff=lfs merge=lfs -text
+data/t_rex.filter_unified.min_entity_16_max_predicate_100.jsonl filter=lfs diff=lfs merge=lfs -text
+data/t_rex.filter_unified.min_entity_16_max_predicate_50.jsonl filter=lfs diff=lfs merge=lfs -text
+data/t_rex.filter_unified.min_entity_4_max_predicate_10.jsonl filter=lfs diff=lfs merge=lfs -text
+data/t_rex.filter_unified.min_entity_4_max_predicate_25.jsonl filter=lfs diff=lfs merge=lfs -text
+data/t_rex.filter_unified.min_entity_12_max_predicate_10.jsonl filter=lfs diff=lfs merge=lfs -text
+data/t_rex.filter_unified.min_entity_12_max_predicate_100.jsonl filter=lfs diff=lfs merge=lfs -text
+data/t_rex.filter_unified.min_entity_12_max_predicate_50.jsonl filter=lfs diff=lfs merge=lfs -text
+data/t_rex.filter_unified.min_entity_4_max_predicate_100.jsonl filter=lfs diff=lfs merge=lfs -text

README.md

@@ -21,7 +21,7 @@ We split the raw T-REX dataset into train/validation/test split by the ratio of
 ### Filtering to Remove Noise
 
 We apply filtering to keep triples with alpha-numeric subject and object, as well as triples with at least either of subject or object is a named-entity.
-After the filtering, we manually remove too vague and noisy predicate, and unify same predicates with different names.
+After the filtering, we manually remove too vague and noisy predicate, and unify same predicates with different names (see the annotation [here](https://huggingface.co/datasets/relbert/t_rex/raw/main/predicate_manual_check.csv)).
 
 | Dataset   | Raw       | Filter    | Unification   |
 |----------:|----------:|----------:|--------------:|
@@ -38,20 +38,21 @@ we choose top-`max predicate` triples based on the frequency of the subject and
 
 - number of triples in each configuration
 
-|   min entity / max predicate |   10 |   25 |    50 |   100 |
-|-----------------------------:|-----:|-----:|------:|------:|
-|                            4 | 4,501 | 9,245 | 15,196 | 23,936 |
-|                            8 | 3,557 | 7,291 | 11,804 | 18,699 |
-|                           12 | 3,132 | 6,346 | 10,155 | 16,115 |
-|                           16 | 2,769 | 5,559 |  9,014 | 14,499 |
+|   min entity / max predicate |   10 |   25 |   50 |   100 |
+|-----------------------------:|-----:|-----:|-----:|------:|
+|                            4 | 1781 | 3961 | 6717 | 10768 |
+|                            8 | 1522 | 3216 | 5286 |  8443 |
+|                           12 | 1301 | 2720 | 4505 |  7191 |
+|                           16 | 1120 | 2389 | 3994 |  6481 |
 
+- number of predicates in different min entity size
 
-|   min entity | predicate |
+|   min entity |   10 |
 |-------------:|-----:|
-|            4 |  516 |
-|            8 |  409 |
-|           12 |  366 |
-|           16 |  321 |
+|            4 |  193 |
+|            8 |  168 |
+|           12 |  146 |
+|           16 |  123 |
 
 - distribution of entities
 

data/stats.data_size.csv

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6a75c050e067579359eef2dd61cfbcc4070654c70aea236153c2463a3191d582
-size 137
+oid sha256:b887e35489f814154dd4fcf5148ed40d43d033e790e38643861007aa6bd8b0c7
+size 131

data/stats.predicate_size.csv

@@ -1,5 +1,5 @@
 min entity,10
-4,516
-8,409
-12,366
-16,321
+4,193
+8,168
+12,146
+16,123

data/t_rex.filter.min_entity_16_max_predicate_100.jsonl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:83df4f92ad5366b1c1b30b1d82d27ba15cbf9fe8656ac25e9253267987774e8d
-size 19047802

data/t_rex.filter.min_entity_16_max_predicate_50.jsonl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:4192b21dd6767564f60d9e4ad5762d92753fd53ba1d0fca3ff051ee5b6926b21
-size 11967293

data/t_rex.filter.min_entity_4_max_predicate_10.jsonl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:063063799b5aaa4c55a8748afad4c5fa76e5eaceda9ed0dc52d9709acb7fc829
-size 6096470

data/t_rex.filter.min_entity_4_max_predicate_100.jsonl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:6ce68333aaf726e68cf9f2f3792f270e993674171a84cf5854b8c5e1760fa777
-size 31419865

data/t_rex.filter.min_entity_4_max_predicate_25.jsonl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:a3028c980fec33c7fb5f4e17951bdc5e1daa86a83f2cf4b5bd6ab6611e1caa9b
-size 12353777

data/t_rex.filter.min_entity_4_max_predicate_50.jsonl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:8fbc0df1edd1c3c317d7f45748deeb608391c64d3de11162664b996f540b5009
-size 20301813

data/t_rex.filter.min_entity_8_max_predicate_10.jsonl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:1f35c4c8a0a1fe68b8538aacc6fbfb60b61724b33a2eec63c3b1d0c621dd33b5
-size 4772164

data/t_rex.filter.min_entity_8_max_predicate_100.jsonl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:5a1bbdc98cabc4c761eaa1b11ee7fb8a6a401c93716c5f06538623bcc38c598e
-size 24795168

data/t_rex.filter.min_entity_8_max_predicate_25.jsonl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:3c6188aa66d44d67c949ab872f22c1039de3f0affff792025465423ac9262d3e
-size 9800900

data/t_rex.filter.min_entity_8_max_predicate_50.jsonl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:62d57cf7448be05fe232100b11baddde6d1fbe48e5c3cb4e4b50abf125fd081d
-size 15780383

data/{t_rex.filter.min_entity_12_max_predicate_10.jsonl → t_rex.filter_unified.min_entity_12_max_predicate_10.jsonl}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6c82acb288dd26a2db2fde6e5cae77bf9dce42b2375aa1250e2bcf101b755360
-size 4206535
+oid sha256:bee332baf23f3b30d97ac34ff35d21daa0bfc44949fb81b5a53260146590852e
+size 1717182

data/{t_rex.filter.min_entity_16_max_predicate_10.jsonl → t_rex.filter_unified.min_entity_12_max_predicate_100.jsonl}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a2d0446d54f35f8dc616499a42f04a088ab84072ad057e0f32a240f12d192599
-size 3748775
+oid sha256:ab2e88b760860317b0033480f0abb15b800b1a8fa14df243b605add2c863e57a
+size 8723923

data/{t_rex.filter.min_entity_12_max_predicate_25.jsonl → t_rex.filter_unified.min_entity_12_max_predicate_25.jsonl}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:83440ded067fb54b368af4b4af4ea1476e407c6131abf8e3885045b1aafe7398
-size 8370852
+oid sha256:a5b773e226a1f4d2b7affb0d017d9f6ac54b7df216d1fdd48a18e0f983163cf4
+size 3415381

data/{t_rex.filter.min_entity_16_max_predicate_25.jsonl → t_rex.filter_unified.min_entity_12_max_predicate_50.jsonl}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:02f2dc1e885f53989a8fa7774bb899ce30d4cf03c1cf945cf9331c5d64a451f6
-size 7364459
+oid sha256:12973871cd44c9441e646dcdae12de370e273dfd8caa8b5710d0bb192f8b56e5
+size 5611692

data/t_rex.filter_unified.min_entity_16_max_predicate_10.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dfcb02689f593cf273b44818fe20ad9a551c7343a9d8196b6f26c05671bc6858
+size 1456921

data/t_rex.filter_unified.min_entity_16_max_predicate_100.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a82011a3a3d874112a6ec09bc0cb014a762189538a4adafbcb64be50d1e791ca
+size 7808966

data/t_rex.filter_unified.min_entity_16_max_predicate_25.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c42756cf46607a62183eb284be0eb1400ae02acc07d39e84b0aca118badac7c
+size 2985297

data/t_rex.filter_unified.min_entity_16_max_predicate_50.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:72033a65fef1890b808e45f9524d4c30dd99480255e155bcb8ad68eb6e847a36
+size 4926634

data/t_rex.filter_unified.min_entity_4_max_predicate_10.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:364307a693fdbdf09fc5db63416294539715d8339484b0077590f8d254ba1ed5
+size 2384750

data/{t_rex.filter.min_entity_12_max_predicate_100.jsonl → t_rex.filter_unified.min_entity_4_max_predicate_100.jsonl}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:54c4e51e6d2f26596009faa44b642f55ab5c7675f1a87ec825441940dd745446
-size 21314377
+oid sha256:a50d482e305403b1a8ff54dbb1fe4fb00f718618191bced35a15c88038960408
+size 13630042

data/t_rex.filter_unified.min_entity_4_max_predicate_25.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c1e0035edda7cdf139f0939cbc04fd55e09eae905ade8971647b069225b823b
+size 5219065

data/t_rex.filter_unified.min_entity_4_max_predicate_50.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d65cf6728fc550108c6b59a313ab93a14461ca636ff08d5d2a4e3ed3b37ddaaf
+size 8669765

data/t_rex.filter_unified.min_entity_8_max_predicate_10.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b64f639231c042eb2b22132b24fd12487ecd705e604b16841bd562c7422733d
+size 2001091

data/{t_rex.filter.min_entity_12_max_predicate_50.jsonl → t_rex.filter_unified.min_entity_8_max_predicate_100.jsonl}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1be72451c6aeed8dc3490703b4b675db9872fbc4b524f71cc8b4fddecaeb0bc3
-size 13536253
+oid sha256:bc844a1e884d23209bd26cf26ee801337e545014dfc532f2d8304d6e59d0b836
+size 10490218

data/t_rex.filter_unified.min_entity_8_max_predicate_25.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b4a117fa4dfc3e7fb26bcc09d48f62c1e3388002caef665fa7d4e101c842eee
+size 4120754

data/t_rex.filter_unified.min_entity_8_max_predicate_50.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f15ad4d9ea3907ee2c8e8e40f5fe06f57aa3fcf77a6af62c633e4a8ff006f96b
+size 6698255

filtering_purify.py

@@ -19,7 +19,7 @@ assert len(parameters_max_p_freq) == 4
 sns.set_theme(style="whitegrid")
 
 # load filtered data
-with open(f"data/t_rex.filter.jsonl") as f:
+with open(f"data/t_rex.filter_unified.jsonl") as f:
     data = Dataset.from_list([json.loads(i) for i in f.read().split('\n') if len(i) > 0])
     df_main = data.to_pandas()
 
@@ -99,7 +99,7 @@ if __name__ == '__main__':
         data_size_full.append(data_size)
         config.append([min_e_freq, max_p_freq])
         # save data
-        with open(f"data/t_rex.filter.min_entity_{min_e_freq}_max_predicate_{max_p_freq}.jsonl", 'w') as f:
+        with open(f"data/t_rex.filter_unified.min_entity_{min_e_freq}_max_predicate_{max_p_freq}.jsonl", 'w') as f:
             f.write('\n'.join([json.dumps(i) for i in new_data]))
 
     # check statistics

t_rex.py

@@ -1,4 +1,6 @@
 import json
+from itertools import product
+
 import datasets
 
 
@@ -17,7 +19,11 @@ _CITATION = """
 
 _HOME_PAGE = "https://github.com/asahi417/relbert"
 _URL = f'https://huggingface.co/datasets/relbert/{_NAME}/resolve/main/data'
+MIN_ENTITY_FREQ = [4, 8, 12, 16]
+MAX_PREDICATE_FREQ = [100, 50, 25, 10]
+
 _TYPES = ["raw", "filter", "filter_unified"]
+# _TYPES += [f"filter_unified.min_entity_{a}_max_predicate_{b}" for a, b in product(MIN_ENTITY_FREQ, MAX_PREDICATE_FREQ)]
 _NON_SPLITS = ["raw", "filter", "filter_unified"]
 _URLS = {i: {str(datasets.Split.TRAIN): [f'{_URL}/t_rex.{i}.jsonl']} if i in _NON_SPLITS else {
     str(datasets.Split.TRAIN): [f'{_URL}/t_rex.{i}.train.jsonl'],