Update spaCy pipeline

.gitattributes

@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+sr_ner_tesla_bcx-any-py3-none-any.whl filter=lfs diff=lfs merge=lfs -text
+transformer/model filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,59 @@
+---
+tags:
+- spacy
+- token-classification
+language:
+- sr
+license: cc-by-sa-3.0
+model-index:
+- name: sr_ner_tesla_bcx
+  results:
+  - task:
+      name: NER
+      type: token-classification
+    metrics:
+    - name: NER Precision
+      type: precision
+      value: 0.957813548
+    - name: NER Recall
+      type: recall
+      value: 0.966107428
+    - name: NER F Score
+      type: f_score
+      value: 0.9619426108
+---
+sr_ner_tesla_bcx is a spaCy model meticulously fine-tuned for Named Entity Recognition in Serbian language texts. This advanced model incorporates a transformer layer based on XLM-R-BERTić, enhancing its analytical capabilities. It is proficient in identifying 7 distinct categories of entities: PERS (persons), ROLE (professions), DEMO (demonyms), ORG (organizations), LOC (locations), WORK (artworks), and EVENT (events). Detailed information about these categories is available in the accompanying table. The development of this model has been made possible through the support of the Science Fund of the Republic of Serbia, under grant #7276, for the project 'Text Embeddings - Serbian Language Applications - TESLA'.
+
+| Feature | Description |
+| --- | --- |
+| **Name** | `sr_ner_tesla_bcx` |
+| **Version** | `1.0.0` |
+| **spaCy** | `>=3.7.2,<3.8.0` |
+| **Default Pipeline** | `transformer`, `ner` |
+| **Components** | `transformer`, `ner` |
+| **Vectors** | 0 keys, 0 unique vectors (0 dimensions) |
+| **Sources** | n/a |
+| **License** | `CC BY-SA 3.0` |
+| **Author** | [Milica Ikonić Nešić, Saša Petalinkar, Mihailo Škorić, Ranka Stanković](https://tesla.rgf.bg.ac.rs/) |
+
+### Label Scheme
+
+<details>
+
+<summary>View label scheme (7 labels for 1 components)</summary>
+
+| Component | Labels |
+| --- | --- |
+| **`ner`** | `DEMO`, `EVENT`, `LOC`, `ORG`, `PERS`, `ROLE`, `WORK` |
+
+</details>
+
+### Accuracy
+
+| Type | Score |
+| --- | --- |
+| `ENTS_F` | 96.19 |
+| `ENTS_P` | 95.78 |
+| `ENTS_R` | 96.61 |
+| `TRANSFORMER_LOSS` | 31553.16 |
+| `NER_LOSS` | 78169.32 |

compare.py

@@ -0,0 +1,45 @@
+import spacy
+from spacy.tokens import DocBin
+import json
+
+def compare_ner_pipelines(binary_file_path, pipeline_names, output_file_path):
+    # Load SpaCy models based on provided pipeline names
+    nlp_pipelines = [spacy.load(name) for name in pipeline_names]
+
+    # Load documents from a binary file
+    doc_bin = DocBin().from_disk(binary_file_path)
+    docs = list(doc_bin.get_docs(nlp_pipelines[0].vocab))  # assuming all models share the same vocab
+
+    # Function to extract entities with their positions
+    def extract_entities(doc):
+        return {(ent.text, ent.start_char, ent.end_char, ent.label_) for ent in doc.ents}
+
+    # Compare entities in each document across all pipelines
+    all_entities_comparison = []
+    for doc in docs:
+        # Include manually annotated entities as the first item in the list
+        entities_per_pipeline = [extract_entities(nlp(doc.text)) for nlp in nlp_pipelines]
+
+        # Find common and unique entities
+        common_entities = set.intersection(*entities_per_pipeline)
+        unique_entities = [ents - common_entities for ents in entities_per_pipeline]
+
+        # Append results for each document
+        all_entities_comparison.append({
+            "document_text": doc.text,
+            "common_entities": list(common_entities),
+            "unique_entities_per_pipeline": {i: list(ents) for i, ents in enumerate(unique_entities)},
+        })
+    # Save the results to a file
+    with open(output_file_path, 'w', encoding="utf-16") as f:
+        json.dump(all_entities_comparison, f, indent=4, ensure_ascii=False)
+
+    print(f"Comparison results saved to {output_file_path}")
+
+# Example usage
+def main():
+    base_path = r"E:\ICIST-2024-models\spacy-tr\spacy-tr"
+    compare_ner_pipelines("SRP19101_1.spacy", [base_path + r"\output20\model-best", base_path + r"\output17\model-best"], "compare_results_nk.json")
+
+if __name__ == "__main__":
+    main()

config.cfg

@@ -0,0 +1,147 @@
+[paths]
+train = "./train.spacy"
+dev = "./dev.spacy"
+vectors = null
+init_tok2vec = null
+
+[system]
+gpu_allocator = "pytorch"
+seed = 0
+
+[nlp]
+lang = "sr"
+pipeline = ["transformer","ner"]
+batch_size = 128
+disabled = []
+before_creation = null
+after_creation = null
+after_pipeline_creation = null
+tokenizer = {"@tokenizers":"spacy.Tokenizer.v1"}
+vectors = {"@vectors":"spacy.Vectors.v1"}
+
+[components]
+
+[components.ner]
+factory = "ner"
+incorrect_spans_key = null
+moves = null
+scorer = {"@scorers":"spacy.ner_scorer.v1"}
+update_with_oracle_cut_size = 100
+
+[components.ner.model]
+@architectures = "spacy.TransitionBasedParser.v2"
+state_type = "ner"
+extra_state_tokens = false
+hidden_width = 64
+maxout_pieces = 2
+use_upper = false
+nO = null
+
+[components.ner.model.tok2vec]
+@architectures = "spacy-transformers.TransformerListener.v1"
+grad_factor = 1.0
+pooling = {"@layers":"reduce_mean.v1"}
+upstream = "*"
+
+[components.transformer]
+factory = "transformer"
+max_batch_items = 4096
+set_extra_annotations = {"@annotation_setters":"spacy-transformers.null_annotation_setter.v1"}
+
+[components.transformer.model]
+@architectures = "spacy-transformers.TransformerModel.v3"
+name = "classla/xlm-r-bertic"
+mixed_precision = false
+
+[components.transformer.model.get_spans]
+@span_getters = "spacy-transformers.strided_spans.v1"
+window = 128
+stride = 96
+
+[components.transformer.model.grad_scaler_config]
+
+[components.transformer.model.tokenizer_config]
+use_fast = true
+
+[components.transformer.model.transformer_config]
+
+[corpora]
+
+[corpora.dev]
+@readers = "spacy.Corpus.v1"
+path = ${paths.dev}
+max_length = 0
+gold_preproc = false
+limit = 0
+augmenter = null
+
+[corpora.train]
+@readers = "spacy.Corpus.v1"
+path = ${paths.train}
+max_length = 0
+gold_preproc = false
+limit = 0
+augmenter = null
+
+[training]
+accumulate_gradient = 3
+dev_corpus = "corpora.dev"
+train_corpus = "corpora.train"
+seed = ${system.seed}
+gpu_allocator = ${system.gpu_allocator}
+dropout = 0.1
+patience = 1600
+max_epochs = 0
+max_steps = 20000
+eval_frequency = 200
+frozen_components = []
+annotating_components = []
+before_to_disk = null
+before_update = null
+
+[training.batcher]
+@batchers = "spacy.batch_by_padded.v1"
+discard_oversize = true
+size = 2000
+buffer = 256
+get_length = null
+
+[training.logger]
+@loggers = "spacy.ConsoleLogger.v1"
+progress_bar = false
+
+[training.optimizer]
+@optimizers = "Adam.v1"
+beta1 = 0.9
+beta2 = 0.999
+L2_is_weight_decay = true
+L2 = 0.01
+grad_clip = 1.0
+use_averages = false
+eps = 0.00000001
+
+[training.optimizer.learn_rate]
+@schedules = "warmup_linear.v1"
+warmup_steps = 250
+total_steps = 20000
+initial_rate = 0.00005
+
+[training.score_weights]
+ents_f = 1.0
+ents_p = 0.0
+ents_r = 0.0
+ents_per_type = null
+
+[pretraining]
+
+[initialize]
+vectors = ${paths.vectors}
+init_tok2vec = ${paths.init_tok2vec}
+vocab_data = null
+lookups = null
+before_init = null
+after_init = null
+
+[initialize.components]
+
+[initialize.tokenizer]

meta.json

@@ -0,0 +1,90 @@
+{
+  "lang":"sr",
+  "name":"ner_tesla_bcx",
+  "version":"1.0.0",
+  "description":"sr_ner_tesla_bcx is a spaCy model meticulously fine-tuned for Named Entity Recognition in Serbian language texts. This advanced model incorporates a transformer layer based on XLM-R-BERTi\u0107, enhancing its analytical capabilities. It is proficient in identifying 7 distinct categories of entities: PERS (persons), ROLE (professions), DEMO (demonyms), ORG (organizations), LOC (locations), WORK (artworks), and EVENT (events). Detailed information about these categories is available in the accompanying table. The development of this model has been made possible through the support of the Science Fund of the Republic of Serbia, under grant #7276, for the project 'Text Embeddings - Serbian Language Applications - TESLA'.",
+  "author":"Milica Ikoni\u0107 Ne\u0161i\u0107, Sa\u0161a Petalinkar, Mihailo \u0160kori\u0107, Ranka Stankovi\u0107",
+  "email":"",
+  "url":"https://tesla.rgf.bg.ac.rs/",
+  "license":"CC BY-SA 3.0",
+  "spacy_version":">=3.7.2,<3.8.0",
+  "spacy_git_version":"a89eae928",
+  "vectors":{
+    "width":0,
+    "vectors":0,
+    "keys":0,
+    "name":null
+  },
+  "labels":{
+    "transformer":[
+
+    ],
+    "ner":[
+      "DEMO",
+      "EVENT",
+      "LOC",
+      "ORG",
+      "PERS",
+      "ROLE",
+      "WORK"
+    ]
+  },
+  "pipeline":[
+    "transformer",
+    "ner"
+  ],
+  "components":[
+    "transformer",
+    "ner"
+  ],
+  "disabled":[
+
+  ],
+  "performance":{
+    "ents_f":0.9619426108,
+    "ents_p":0.957813548,
+    "ents_r":0.966107428,
+    "ents_per_type":{
+      "ROLE":{
+        "p":0.8736710444,
+        "r":0.9172685489,
+        "f":0.8949391416
+      },
+      "PERS":{
+        "p":0.9903689806,
+        "r":0.9903689806,
+        "f":0.9903689806
+      },
+      "LOC":{
+        "p":0.9556978233,
+        "r":0.983658408,
+        "f":0.9694765554
+      },
+      "DEMO":{
+        "p":0.915407855,
+        "r":0.9543307087,
+        "f":0.934464148
+      },
+      "ORG":{
+        "p":0.8515742129,
+        "r":0.7759562842,
+        "f":0.8120085776
+      },
+      "WORK":{
+        "p":0.76,
+        "r":0.5352112676,
+        "f":0.6280991736
+      },
+      "EVENT":{
+        "p":0.6875,
+        "r":0.6875,
+        "f":0.6875
+      }
+    },
+    "transformer_loss":315.5316300207,
+    "ner_loss":781.6932183095
+  },
+  "requirements":[
+    "spacy-transformers>=1.3.4,<1.4.0"
+  ]
+}

ner/cfg

@@ -0,0 +1,13 @@
+{
+  "moves":null,
+  "update_with_oracle_cut_size":100,
+  "multitasks":[
+
+  ],
+  "min_action_freq":1,
+  "learn_tokens":false,
+  "beam_width":1,
+  "beam_density":0.0,
+  "beam_update_prob":0.0,
+  "incorrect_spans_key":null
+}

ner/moves

@@ -0,0 +1 @@
+��moves�
�{"0":{},"1":{"PERS":66081,"LOC":35152,"ROLE":14259,"ORG":10504,"DEMO":5087,"WORK":973,"EVENT":546},"2":{"PERS":66081,"LOC":35152,"ROLE":14259,"ORG":10504,"DEMO":5087,"WORK":973,"EVENT":546},"3":{"PERS":66081,"LOC":35152,"ROLE":14259,"ORG":10504,"DEMO":5087,"WORK":973,"EVENT":546},"4":{"PERS":66081,"LOC":35152,"ROLE":14259,"ORG":10504,"DEMO":5087,"WORK":973,"EVENT":546,"":1},"5":{"":1}}�cfg��neg_key�

sr_ner_tesla_bcx-any-py3-none-any.whl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1700bc3ddcbdc55cffecc1bb305e2827caf2245d0509e3889c8140b020959357
+size 2085012613

transformer/cfg

@@ -0,0 +1,3 @@
+{
+  "max_batch_items":4096
+}

transformer/model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90c7ca694d8c4ca3b6500046f95a60ec8a72f7a704f5750d5e9df8fa4966ca6f
+size 2261874822

vocab/key2row

@@ -0,0 +1 @@
+�

vocab/lookups.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76be8b528d0075f7aae98d6fa57a6d3c83ae480a8469e668d7b0af968995ac71
+size 1

vocab/vectors.cfg

@@ -0,0 +1,3 @@
+{
+  "mode":"default"
+}