Ajout de 4 model

model/HuggingFace/BartLargeMnli.py

-# https://huggingface.co/facebook/bart-large-mnli
-
-from transformers import pipeline
-
-classifier = pipeline("zero-shot-classification", model="facebook/bart-large-mnli")
-candidate_labels = [
-        "Diabetes",
-        "Cancer",
-        "Chronic respiratory disease",
-        "Cardiovascular diseases",
-        "Mental Health",
-        "Diabetes type 1",
-        "Diabetes type 2"
-        ]
-
-def classify(sequence):
-    results = classifier(sequence, candidate_labels)
-
-    return results
-
-    # print(f"Sequence: {sequence_to_classify}")
-    # print(f"Labels: {results["labels"]}")
-    # print(f"Scores: {results["scores"]}")

model/HuggingFace/BioMedBERT.py

-# https://huggingface.co/microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract
-
-from transformers import pipeline
-
-classifier = pipeline("zero-shot-classification", model="microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract")
-candidate_labels = [
-        "Diabetes",
-        "Cancer",
-        "Chronic respiratory disease",
-        "Cardiovascular diseases",
-        "Mental Health",
-        "Diabetes type 1",
-        "Diabetes type 2"
-        ]
-
-def classify(sequence):
-    results = classifier(sequence, candidate_labels)
-
-    return results
-
-    # print(f"Sequence: {sequence_to_classify}")
-    # print(f"Labels: {results["labels"]}")
-    # print(f"Scores: {results["scores"]}")

model/HuggingFace/bge_m3_zeroshot_v2.py

-# https://huggingface.co/MoritzLaurer/bge-m3-zeroshot-v2.0
-
-from transformers import pipeline
-
-classifier = pipeline("zero-shot-classification", model="MoritzLaurer/bge-m3-zeroshot-v2.0")
-candidate_labels = [
-        "Diabetes",
-        "Cancer",
-        "Chronic respiratory disease",
-        "Cardiovascular diseases",
-        "Mental Health",
-        "Diabetes type 1",
-        "Diabetes type 2"
-        ]
-
-def classify(sequence):
-    results = classifier(sequence, candidate_labels)
-
-    return results
-
-    # print(f"Sequence: {sequence_to_classify}")
-    # print(f"Labels: {results["labels"]}")
-    # print(f"Scores: {results["scores"]}")

model/HuggingFace/zero_shot_classification.py

+from transformers import pipeline
+
+LABELS = [
+    "Diabetes",
+    "Cancer",
+    "Chronic respiratory disease",
+    "Cardiovascular diseases",
+    "Mental Health",
+    "Diabetes type 1",
+    "Diabetes type 2"
+]
+
+MODELS = [
+    "facebook/bart-large-mnli", # https://huggingface.co/facebook/bart-large-mnli
+    "MoritzLaurer/bge-m3-zeroshot-v2.0", # https://huggingface.co/MoritzLaurer/bge-m3-zeroshot-v2.0
+    "MoritzLaurer/DeBERTa-v3-base-mnli-fever-anli",
+    "MoritzLaurer/deberta-v3-base-zeroshot-v1.1-all-33",
+    "MoritzLaurer/multilingual-MiniLMv2-L6-mnli-xnli",
+    "microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract", # https://huggingface.co/microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract
+]
+
+def create_classifier(model = MODELS[0]):
+    return pipeline("zero-shot-classification", model=model)
+
+def classify(classifier, sequence, lables = LABELS, debug = False):
+    results = classifier(sequence, lables)
+
+    if debug:
+        print(f"Sequence: {sequence}")
+        print(f"Labels: {results["labels"]}")
+        print(f"Scores: {results["scores"]}")
+
+    return results

model/results/bge_m3_zeroshot_v2.txt → model/results/MoritzLaurer-bge_m3_zeroshot_v2_0.txt

@@ -502,7 +502,7 @@ Labels: ['Cardiovascular diseases', 'Diabetes', 'Diabetes type 2', 'Cancer', 'Di
 Scores: [0.38244596123695374, 0.23532404005527496, 0.12610754370689392, 0.08088988065719604, 0.07161928713321686, 0.055826857686042786, 0.04778643324971199]
 Selected labels: []
 ---------------------------------
-Time to classify all articles: 374.61803555488586 seconds
+Time to classify all articles: 380.2531027793884 seconds
 Good classification: 12/75 ( 16.0% )
 At least one good per entrie: 12/63 ( 19.05% )
 Mean score: 0.7297625946998596

model/results/BartLargeMnli.txt → model/results/facebook-bart_large_mnli.txt

@@ -502,7 +502,7 @@ Labels: ['Cardiovascular diseases', 'Diabetes type 2', 'Diabetes', 'Chronic resp
 Scores: [0.34633418917655945, 0.2441960871219635, 0.09421524405479431, 0.09026889503002167, 0.08233360201120377, 0.07539966702461243, 0.06725231558084488]
 Selected labels: []
 ---------------------------------
-Time to classify all articles: 418.86657190322876 seconds
+Time to classify all articles: 420.01833271980286 seconds
 Good classification: 31/75 ( 41.33% )
 At least one good per entrie: 31/63 ( 49.21% )
 Mean score: 0.797397155972088

model/test_model.py

@@ -8,9 +8,7 @@ from api.parser.jsonParser import parseJsonFile
 import time
 import statistics
 
-# from model.HuggingFace.BartLargeMnli import classify
-# from model.HuggingFace.BioMedBERT import classify
-from model.HuggingFace.bge_m3_zeroshot_v2 import classify
+from model.HuggingFace.zero_shot_classification import create_classifier, classify, MODELS
 
 TRESHOLD = 0.6
 
@@ -20,6 +18,9 @@ KEYWORDS = [
         "Diabetes type 2"
         ]
 
+model = MODELS[6]
+classifier = create_classifier(model)
+
 def predict(article, file):
     pmid = article["PMID"]
     title = article["Title"]
@@ -34,11 +35,10 @@ def predict(article, file):
     print(f"Predictions: {pred}", file=file)
     print(f"MeshTerm: {meshTerms}", file=file)
 
-    return classify(title + " \n " + articleTitle + " \n " + abstract)
+    return classify(classifier, title + " \n " + articleTitle + " \n " + abstract)
 
-# with open("./results/BartLargeMnli.txt", "w") as file:
-# with open("./results/BiomedBERT.txt", "w") as file:
-with open("./results/bge_m3_zeroshot_v2.txt", "w") as file:
+filename = model.replace(" ", "_").replace("-", "_").replace(".", "_").replace("/", "-")
+with open(f"./results/{filename}.txt", "w") as file:
 
     predictions = []
     selected_scores = []