LunaLan07 commited on
Commit
0b778f7
·
verified ·
1 Parent(s): ae7ed46

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +11 -11
README.md CHANGED
@@ -51,29 +51,29 @@ BioHiCL aligns:
51
  ```python
52
  from beir import util
53
  from beir.datasets.data_loader import GenericDataLoader
54
- from sentence_transformers import SentenceTransformer
55
  from beir.retrieval.search.dense import DenseRetrievalExactSearch
56
  from beir.retrieval.evaluation import EvaluateRetrieval
57
 
58
- # Dataset
 
59
  dataset = "scifact"
60
- url = "https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/scifact.zip"
61
 
62
  data_path = util.download_and_unzip(url, "datasets")
63
  corpus, queries, qrels = GenericDataLoader(data_path).load(split="test")
64
 
65
- # Model
66
  model_name = "LunaLan07/BioHiCL-base"
67
- model = SentenceTransformer(model_name)
68
 
69
- # Retrieval
70
  retriever = DenseRetrievalExactSearch(model, batch_size=16)
71
- results = retriever.search(corpus, queries, top_k=10, score_function="cos_sim")
72
 
73
- # Evaluation
74
- ndcg, _map, recall, precision = EvaluateRetrieval.evaluate(
75
- qrels, results, k_values=[1, 3, 5, 10]
76
- )
 
 
77
 
78
  ```
79
 
 
51
  ```python
52
  from beir import util
53
  from beir.datasets.data_loader import GenericDataLoader
54
+ from beir.retrieval.models import SentenceBERT
55
  from beir.retrieval.search.dense import DenseRetrievalExactSearch
56
  from beir.retrieval.evaluation import EvaluateRetrieval
57
 
58
+
59
+ # 1. Download load the SciFact dataset
60
  dataset = "scifact"
61
+ url = "https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/" + dataset + ".zip"
62
 
63
  data_path = util.download_and_unzip(url, "datasets")
64
  corpus, queries, qrels = GenericDataLoader(data_path).load(split="test")
65
 
 
66
  model_name = "LunaLan07/BioHiCL-base"
67
+ model = SentenceBERT(model_name)
68
 
 
69
  retriever = DenseRetrievalExactSearch(model, batch_size=16)
 
70
 
71
+ top_k = 10 # top 10 documents per query
72
+ results = retriever.search(corpus, queries, top_k=top_k, score_function="cos_sim")
73
+
74
+ k_values = [1, 3, 5, 10]
75
+ ndcg, _map, recall, precision = EvaluateRetrieval.evaluate(qrels, results, k_values=k_values)
76
+
77
 
78
  ```
79