Upload model 0.8.1

LICENSE.md

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2023 CodeNLP Michał Marcińczuk
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -1,3 +1,23 @@
----
-license: mit
----
+## About
+
+A transformer-based model for named entity recognition for Polish.
+The model was trained on the train part of the KPWr corpus to recognize
+82 categories of NE.
+
+## Evaluation summary
+
+                           precision    recall  f1-score   support
+micro avg                     0.7803    0.8033    0.7916     17711
+macro avg                     0.7921    0.8033    0.7932     17711
+
+Detailed results can be found in the test_results.txt file.
+
+
+## Author
+
+Michał Marcińczuk <marcinczuk@gmail.com>
+
+
+## License
+
+[MIT](https://choosealicense.com/licenses/mit/

config.json

@@ -0,0 +1,31 @@
+{
+  "_name_or_path": "allegro/herbert-large-cased",
+  "architectures": [
+    "BertModel"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "directionality": "bidi",
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 514,
+  "model_type": "bert",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 24,
+  "pad_token_id": 1,
+  "pooler_fc_size": 768,
+  "pooler_num_attention_heads": 12,
+  "pooler_num_fc_layers": 3,
+  "pooler_size_per_head": 128,
+  "pooler_type": "first_token_transform",
+  "position_embedding_type": "absolute",
+  "tokenizer_class": "HerbertTokenizerFast",
+  "transformers_version": "4.17.0",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 50000
+}

pdn2_config.yml

@@ -0,0 +1,171 @@
+!!python/object:poldeepner2.model.hf_for_token_calssification.Pdn2ModelConfiguration
+device: cuda:0
+dropout_p: 0.2
+head_init_range: 0.04
+hidden_size: 1024
+label_ignore_idx: 0
+labels:
+- B-nam_adj
+- B-nam_adj_city
+- B-nam_adj_country
+- B-nam_adj_person
+- B-nam_eve
+- B-nam_eve_human
+- B-nam_eve_human_cultural
+- B-nam_eve_human_holiday
+- B-nam_eve_human_sport
+- B-nam_fac_bridge
+- B-nam_fac_goe
+- B-nam_fac_goe_stop
+- B-nam_fac_park
+- B-nam_fac_road
+- B-nam_fac_square
+- B-nam_fac_system
+- B-nam_liv_animal
+- B-nam_liv_character
+- B-nam_liv_god
+- B-nam_liv_habitant
+- B-nam_liv_person
+- B-nam_loc
+- B-nam_loc_astronomical
+- B-nam_loc_country_region
+- B-nam_loc_gpe_admin1
+- B-nam_loc_gpe_admin2
+- B-nam_loc_gpe_admin3
+- B-nam_loc_gpe_city
+- B-nam_loc_gpe_conurbation
+- B-nam_loc_gpe_country
+- B-nam_loc_gpe_district
+- B-nam_loc_gpe_subdivision
+- B-nam_loc_historical_region
+- B-nam_loc_hydronym
+- B-nam_loc_hydronym_lake
+- B-nam_loc_hydronym_ocean
+- B-nam_loc_hydronym_river
+- B-nam_loc_hydronym_sea
+- B-nam_loc_land
+- B-nam_loc_land_continent
+- B-nam_loc_land_island
+- B-nam_loc_land_mountain
+- B-nam_loc_land_peak
+- B-nam_loc_land_region
+- B-nam_num_house
+- B-nam_num_phone
+- B-nam_org_company
+- B-nam_org_group
+- B-nam_org_group_band
+- B-nam_org_group_team
+- B-nam_org_institution
+- B-nam_org_nation
+- B-nam_org_organization
+- B-nam_org_organization_sub
+- B-nam_org_political_party
+- B-nam_oth
+- B-nam_oth_currency
+- B-nam_oth_data_format
+- B-nam_oth_license
+- B-nam_oth_position
+- B-nam_oth_tech
+- B-nam_oth_www
+- B-nam_pro
+- B-nam_pro_award
+- B-nam_pro_brand
+- B-nam_pro_media
+- B-nam_pro_media_periodic
+- B-nam_pro_media_radio
+- B-nam_pro_media_tv
+- B-nam_pro_media_web
+- B-nam_pro_model_car
+- B-nam_pro_software
+- B-nam_pro_software_game
+- B-nam_pro_title
+- B-nam_pro_title_album
+- B-nam_pro_title_article
+- B-nam_pro_title_book
+- B-nam_pro_title_document
+- B-nam_pro_title_song
+- B-nam_pro_title_treaty
+- B-nam_pro_title_tv
+- B-nam_pro_vehicle
+- I-nam_adj_country
+- I-nam_eve
+- I-nam_eve_human
+- I-nam_eve_human_cultural
+- I-nam_eve_human_holiday
+- I-nam_eve_human_sport
+- I-nam_fac_bridge
+- I-nam_fac_goe
+- I-nam_fac_goe_stop
+- I-nam_fac_park
+- I-nam_fac_road
+- I-nam_fac_square
+- I-nam_fac_system
+- I-nam_liv_animal
+- I-nam_liv_character
+- I-nam_liv_god
+- I-nam_liv_person
+- I-nam_loc
+- I-nam_loc_astronomical
+- I-nam_loc_country_region
+- I-nam_loc_gpe_admin1
+- I-nam_loc_gpe_admin2
+- I-nam_loc_gpe_admin3
+- I-nam_loc_gpe_city
+- I-nam_loc_gpe_conurbation
+- I-nam_loc_gpe_country
+- I-nam_loc_gpe_district
+- I-nam_loc_gpe_subdivision
+- I-nam_loc_historical_region
+- I-nam_loc_hydronym
+- I-nam_loc_hydronym_lake
+- I-nam_loc_hydronym_ocean
+- I-nam_loc_hydronym_river
+- I-nam_loc_hydronym_sea
+- I-nam_loc_land
+- I-nam_loc_land_continent
+- I-nam_loc_land_island
+- I-nam_loc_land_mountain
+- I-nam_loc_land_peak
+- I-nam_loc_land_region
+- I-nam_num_house
+- I-nam_num_phone
+- I-nam_org_company
+- I-nam_org_group
+- I-nam_org_group_band
+- I-nam_org_group_team
+- I-nam_org_institution
+- I-nam_org_nation
+- I-nam_org_organization
+- I-nam_org_organization_sub
+- I-nam_org_political_party
+- I-nam_oth
+- I-nam_oth_currency
+- I-nam_oth_data_format
+- I-nam_oth_license
+- I-nam_oth_position
+- I-nam_oth_tech
+- I-nam_oth_www
+- I-nam_pro
+- I-nam_pro_award
+- I-nam_pro_brand
+- I-nam_pro_media
+- I-nam_pro_media_periodic
+- I-nam_pro_media_radio
+- I-nam_pro_media_tv
+- I-nam_pro_media_web
+- I-nam_pro_model_car
+- I-nam_pro_software
+- I-nam_pro_software_game
+- I-nam_pro_title
+- I-nam_pro_title_album
+- I-nam_pro_title_article
+- I-nam_pro_title_book
+- I-nam_pro_title_document
+- I-nam_pro_title_song
+- I-nam_pro_title_treaty
+- I-nam_pro_title_tv
+- I-nam_pro_vehicle
+- O
+max_seq_length: 256
+seed: 101
+sequence_generator: context-window

pdn2_eval_kpwr_test.txt

@@ -0,0 +1,82 @@
+                           precision    recall  f1-score   support
+
+nam_adj                       0.5741    0.5962    0.5849        52
+nam_adj_city                  0.8261    0.9048    0.8636        42
+nam_adj_country               0.7538    0.8855    0.8144       166
+nam_adj_person                1.0000    0.7778    0.8750        18
+nam_eve                       1.0000    0.8750    0.9333         8
+nam_eve_human                 0.4337    0.4615    0.4472        78
+nam_eve_human_cultural        0.3529    0.2727    0.3077        22
+nam_eve_human_holiday         0.5455    0.6667    0.6000         9
+nam_eve_human_sport           0.6230    0.6909    0.6552        55
+nam_fac_bridge                0.3750    0.7500    0.5000         4
+nam_fac_goe                   0.6071    0.5312    0.5667        64
+nam_fac_goe_stop              0.6000    0.7500    0.6667         4
+nam_fac_park                  0.5556    0.5000    0.5263        10
+nam_fac_road                  0.9100    0.9579    0.9333        95
+nam_fac_square                0.8000    0.6667    0.7273         6
+nam_fac_system                0.8235    0.5385    0.6512        26
+nam_liv_animal                0.6429    0.8182    0.7200        11
+nam_liv_god                   0.8205    0.9143    0.8649        35
+nam_liv_habitant              0.3750    0.4286    0.4000         7
+nam_liv_person                0.9503    0.9663    0.9582       949
+nam_loc                       0.0000    0.0000    0.0000         4
+nam_loc_country_region        0.0667    0.2500    0.1053         4
+nam_loc_gpe_admin1            0.7867    0.9219    0.8489        64
+nam_loc_gpe_admin2            0.8000    0.6667    0.7273        36
+nam_loc_gpe_admin3            0.8140    0.7447    0.7778        47
+nam_loc_gpe_city              0.9032    0.8970    0.9001       437
+nam_loc_gpe_country           0.9536    0.9776    0.9654       357
+nam_loc_gpe_district          0.2439    0.5556    0.3390        18
+nam_loc_gpe_subdivision       0.7143    0.1923    0.3030        26
+nam_loc_historical_region     0.8235    0.6364    0.7179        22
+nam_loc_hydronym              0.0000    0.0000    0.0000         1
+nam_loc_hydronym_lake         1.0000    0.5000    0.6667         2
+nam_loc_hydronym_ocean        1.0000    1.0000    1.0000         1
+nam_loc_hydronym_river        0.8627    0.8627    0.8627        51
+nam_loc_hydronym_sea          1.0000    1.0000    1.0000         3
+nam_loc_land                  0.0000    0.0000    0.0000         2
+nam_loc_land_continent        0.9118    0.9688    0.9394        32
+nam_loc_land_island           1.0000    0.5455    0.7059        11
+nam_loc_land_mountain         0.8571    0.6667    0.7500         9
+nam_loc_land_region           0.5455    0.5455    0.5455        11
+nam_num_house                 0.8462    1.0000    0.9167        11
+nam_num_phone                 1.0000    0.5000    0.6667         2
+nam_org_company               0.6512    0.7368    0.6914        76
+nam_org_group                 0.3571    0.2778    0.3125        18
+nam_org_group_band            0.7895    0.7895    0.7895        19
+nam_org_group_team            0.9412    0.9664    0.9536       149
+nam_org_institution           0.7434    0.7406    0.7420       266
+nam_org_nation                0.9000    0.7778    0.8344        81
+nam_org_organization          0.7754    0.7439    0.7593       246
+nam_org_organization_sub      0.5000    0.3333    0.4000         3
+nam_org_political_party       0.8871    0.9483    0.9167        58
+nam_oth                       0.0732    0.1364    0.0952        22
+nam_oth_currency              0.9362    0.8627    0.8980        51
+nam_oth_data_format           0.0000    0.0000    0.0000        10
+nam_oth_license               0.6429    0.8182    0.7200        11
+nam_oth_position              0.3333    0.7000    0.4516        10
+nam_oth_tech                  0.5000    0.5738    0.5344        61
+nam_oth_www                   0.7500    0.1500    0.2500        20
+nam_pro                       0.0000    0.0000    0.0000         2
+nam_pro_award                 0.7600    0.8261    0.7917        23
+nam_pro_brand                 0.5435    0.5435    0.5435        46
+nam_pro_media                 0.0000    0.0000    0.0000         8
+nam_pro_media_periodic        0.8280    0.9167    0.8701        84
+nam_pro_media_radio           0.5000    0.6667    0.5714         3
+nam_pro_media_tv              0.4667    1.0000    0.6364         7
+nam_pro_media_web             0.3836    0.7000    0.4956        40
+nam_pro_model_car             0.8571    0.9231    0.8889        26
+nam_pro_software              0.7381    0.6392    0.6851        97
+nam_pro_software_game         0.0000    0.0000    0.0000         3
+nam_pro_title                 0.4545    0.5714    0.5063        35
+nam_pro_title_album           0.6667    0.8571    0.7500         7
+nam_pro_title_book            0.2500    0.1818    0.2105        11
+nam_pro_title_document        0.5826    0.7614    0.6601        88
+nam_pro_title_song            0.6000    0.4286    0.5000         7
+nam_pro_title_treaty          0.0000    0.0000    0.0000         2
+nam_pro_title_tv              0.8824    0.6250    0.7317        24
+nam_pro_vehicle               0.0000    0.0000    0.0000         4
+
+micro avg                     0.7905    0.8158    0.8029      4430
+macro avg                     0.8044    0.8158    0.8052      4430

pdn2_metadata.yml

@@ -0,0 +1,6 @@
+author: Michał Marcińczuk, CodeNLP
+contact: marcinczuk@gmail.com
+description: Model trained on the KPWr corpus recognizes 82 types of nested entities.
+reference: https://www.sciencedirect.com/science/article/pii/S1877050921015179
+score: F1=80.29 on the KPWr test subset (micro avg, strict evaluation).
+ 

pdn2_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7cb4fc4bd740196db90ab184b7aad0699dacc12628f5cdf29ab143bd39f3ff35
+size 1425385193

special_tokens_map.json

@@ -0,0 +1 @@
+{"bos_token": "<s>", "unk_token": "<unk>", "sep_token": "</s>", "pad_token": "<pad>", "cls_token": "<s>", "mask_token": "<mask>"}

test_results.txt

@@ -0,0 +1,82 @@
+                           precision    recall  f1-score   support
+
+nam_adj                       0.6230    0.5721    0.5965       208
+nam_adj_city                  0.8343    0.8690    0.8513       168
+nam_adj_country               0.7543    0.8554    0.8017       664
+nam_adj_person                0.9825    0.7778    0.8682        72
+nam_eve                       1.0000    0.8438    0.9153        32
+nam_eve_human                 0.3813    0.4583    0.4163       312
+nam_eve_human_cultural        0.3148    0.1932    0.2394        88
+nam_eve_human_holiday         0.5349    0.6389    0.5823        36
+nam_eve_human_sport           0.6173    0.6818    0.6479       220
+nam_fac_bridge                0.5217    0.7500    0.6154        16
+nam_fac_goe                   0.5983    0.5469    0.5714       256
+nam_fac_goe_stop              0.5000    0.6250    0.5556        16
+nam_fac_park                  0.6111    0.5500    0.5789        40
+nam_fac_road                  0.9333    0.9579    0.9455       380
+nam_fac_square                0.8421    0.6667    0.7442        24
+nam_fac_system                0.7647    0.5000    0.6047       104
+nam_liv_animal                0.5106    0.5455    0.5275        44
+nam_liv_god                   0.7785    0.8786    0.8255       140
+nam_liv_habitant              0.4074    0.3929    0.4000        28
+nam_liv_person                0.9476    0.9629    0.9552      3796
+nam_loc                       0.0185    0.0625    0.0286        16
+nam_loc_country_region        0.0926    0.3125    0.1429        16
+nam_loc_gpe_admin1            0.7607    0.9062    0.8271       256
+nam_loc_gpe_admin2            0.8091    0.6181    0.7008       144
+nam_loc_gpe_admin3            0.8225    0.7394    0.7787       188
+nam_loc_gpe_city              0.8915    0.9022    0.8968      1748
+nam_loc_gpe_country           0.9438    0.9643    0.9539      1428
+nam_loc_gpe_district          0.2198    0.5556    0.3150        72
+nam_loc_gpe_subdivision       0.6452    0.1923    0.2963       104
+nam_loc_historical_region     0.8333    0.6250    0.7143        88
+nam_loc_hydronym              0.0000    0.0000    0.0000         4
+nam_loc_hydronym_lake         1.0000    0.5000    0.6667         8
+nam_loc_hydronym_ocean        1.0000    1.0000    1.0000         4
+nam_loc_hydronym_river        0.8495    0.8578    0.8537       204
+nam_loc_hydronym_sea          0.8571    1.0000    0.9231        12
+nam_loc_land                  0.0000    0.0000    0.0000         8
+nam_loc_land_continent        0.9248    0.9609    0.9425       128
+nam_loc_land_island           0.6667    0.4545    0.5405        44
+nam_loc_land_mountain         0.7692    0.5556    0.6452        36
+nam_loc_land_region           0.5476    0.5227    0.5349        44
+nam_num_house                 0.8627    1.0000    0.9263        44
+nam_num_phone                 1.0000    0.6250    0.7692         8
+nam_org_company               0.6393    0.7171    0.6760       304
+nam_org_group                 0.3409    0.2083    0.2586        72
+nam_org_group_band            0.7922    0.8026    0.7974        76
+nam_org_group_team            0.9329    0.9564    0.9445       596
+nam_org_institution           0.7062    0.7500    0.7274      1064
+nam_org_nation                0.8737    0.7685    0.8177       324
+nam_org_organization          0.7511    0.7175    0.7339       984
+nam_org_organization_sub      0.1538    0.1667    0.1600        12
+nam_org_political_party       0.8810    0.9569    0.9174       232
+nam_oth                       0.0805    0.1364    0.1013        88
+nam_oth_currency              0.9396    0.8382    0.8860       204
+nam_oth_data_format           0.0000    0.0000    0.0000        40
+nam_oth_license               0.7551    0.8409    0.7957        44
+nam_oth_position              0.3766    0.7250    0.4957        40
+nam_oth_tech                  0.5094    0.5574    0.5323       244
+nam_oth_www                   0.5556    0.1250    0.2041        80
+nam_pro                       0.0000    0.0000    0.0000         8
+nam_pro_award                 0.6782    0.6413    0.6592        92
+nam_pro_brand                 0.5202    0.5598    0.5393       184
+nam_pro_media                 0.0000    0.0000    0.0000        32
+nam_pro_media_periodic        0.8149    0.8886    0.8501       332
+nam_pro_media_radio           0.5556    0.8333    0.6667        12
+nam_pro_media_tv              0.4179    1.0000    0.5895        28
+nam_pro_media_web             0.3695    0.6813    0.4791       160
+nam_pro_model_car             0.8246    0.9038    0.8624       104
+nam_pro_software              0.7430    0.6186    0.6751       388
+nam_pro_software_game         0.0000    0.0000    0.0000        12
+nam_pro_title                 0.5063    0.5714    0.5369       140
+nam_pro_title_album           0.6190    0.9286    0.7429        28
+nam_pro_title_book            0.3103    0.2045    0.2466        44
+nam_pro_title_document        0.5137    0.6484    0.5732       347
+nam_pro_title_song            0.7059    0.4286    0.5333        28
+nam_pro_title_treaty          0.0714    0.1250    0.0909         8
+nam_pro_title_tv              0.8636    0.5938    0.7037        96
+nam_pro_vehicle               0.0000    0.0000    0.0000        16
+
+micro avg                     0.7803    0.8033    0.7916     17711
+macro avg                     0.7921    0.8033    0.7932     17711

tokenizer_config.json

@@ -0,0 +1 @@
+{"cls_token": "<s>", "unk_token": "<unk>", "pad_token": "<pad>", "mask_token": "<mask>", "sep_token": "</s>", "do_lowercase_and_remove_accent": false, "bos_token": "<s>", "additional_special_tokens": [], "model_max_length": 512, "special_tokens_map_file": "/home/czuk/.cache/huggingface/transformers/7e8fe8852a1ff7e03195cb41fac16af837f8c14a34a61850b02a7395eb294f00.b8e113717eb1828d09e47de853cf49c8fad05ebdce24df2614cd942dc23e2a77", "name_or_path": "allegro/herbert-large-cased", "lang2id": null, "id2lang": null, "tokenizer_class": "HerbertTokenizer"}