NER and trial task

tutorials/llm_clinical_trials/configs/ner_openai.cfg

@@ -0,0 +1,17 @@
+[nlp]
+lang = "en"
+pipeline = ["llm"]
+batch_size = 128
+
+[components]
+
+[components.llm]
+factory = "llm"
+
+[components.llm.model]
+@llm_models = "spacy.GPT-3-5.v1"
+name = "gpt-3.5-turbo"
+
+[components.llm.task]
+@llm_tasks = "spacy.NER.v2"
+labels = "Drug,Dose"

tutorials/llm_clinical_trials/project.yml

@@ -2,7 +2,8 @@ title: 'Clinical trial results extraction with LLMs'
 description: "Using an LLM in a spaCy pipeline to extract patient groups, treatments and outcomes in clinical trials."
 
 vars:
-  config: "config_openai.cfg"
+  ner_config: "ner_openai.cfg"
+  trial_config: "trial_openai.cfg"
   pmid: 27144689
 
 # These are the directories that the project needs.
@@ -16,10 +17,18 @@ assets:
 
 # Project commands
 commands:
+  - name: zero_shot_ner
+    help: "Run an LLM pipeline with zero-shot NER"
+    script:
+      - "python ./scripts/visualise_entities.py ${vars.pmid} ./configs/${vars.ner_config}"
+    deps:
+      - "assets/${vars.pmid}.txt"
+      - "configs/${vars.ner_config}"
+
   - name: summarize
     help: "Run the LLM-powered spaCy pipeline"
     script:
-      - "python ./scripts/run_pipeline.py ${vars.pmid} ./configs/${vars.config}"
+      - "python ./scripts/extract_results.py ${vars.pmid} ./configs/${vars.trial_config}"
     deps:
       - "assets/${vars.pmid}.txt"
-      - "configs/${vars.config}"
+      - "configs/${vars.trial_config}"

tutorials/llm_clinical_trials/scripts/visualise_entities.py

@@ -0,0 +1,21 @@
+from pathlib import Path
+
+import typer
+from input_reader import read_trial
+from spacy import displacy
+from spacy_llm.util import assemble
+from wasabi import msg
+
+
+def visualise_entities(pmid: int, config_path: Path, verbose: bool = False):
+    msg.text(f"Processing PMID {pmid}", show=verbose)
+    msg.text(f"Loading config from {config_path}", show=verbose)
+    text = read_trial(pmid, verbose=verbose)
+    nlp = assemble(config_path)
+    doc = nlp(text)
+    options = {"ents": ["Drug", "Dose"], "colors": {"Drug": "pink", "Dose": "orange"}}
+    displacy.serve(doc, style="ent", options=options)
+
+
+if __name__ == "__main__":
+    typer.run(visualise_entities)