Duplicate from rag-datasets/rag-mini-bioasq

Co-authored-by: Till Wenke <tillwenke@users.noreply.huggingface.co>

.gitattributes

@@ -0,0 +1,57 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.lz4 filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz 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
+# Audio files - uncompressed
+*.pcm filter=lfs diff=lfs merge=lfs -text
+*.sam filter=lfs diff=lfs merge=lfs -text
+*.raw filter=lfs diff=lfs merge=lfs -text
+# Audio files - compressed
+*.aac filter=lfs diff=lfs merge=lfs -text
+*.flac filter=lfs diff=lfs merge=lfs -text
+*.mp3 filter=lfs diff=lfs merge=lfs -text
+*.ogg filter=lfs diff=lfs merge=lfs -text
+*.wav filter=lfs diff=lfs merge=lfs -text
+# Image files - uncompressed
+*.bmp filter=lfs diff=lfs merge=lfs -text
+*.gif filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+*.tiff filter=lfs diff=lfs merge=lfs -text
+# Image files - compressed
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text
+*.webp filter=lfs diff=lfs merge=lfs -text
+# custom
+raw_data/** filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,2 @@
+/env
+credentials.json

README.md

@@ -0,0 +1,29 @@
+---
+license: cc-by-2.5
+task_categories:
+- question-answering
+- sentence-similarity
+language:
+- en
+tags:
+- rag
+- dpr
+- information-retrieval
+- question-answering
+- biomedical
+configs:
+- config_name: text-corpus
+  data_files: 
+  - split: passages
+    path: "data/passages.parquet/*"
+- config_name: question-answer-passages
+  data_files: 
+  - split: test
+    path: "data/test.parquet/*"
+---
+
+See [here](https://huggingface.co/datasets/enelpol/rag-mini-bioasq) for an updated version without nans in text-corpus.
+
+[In this huggingface discussion](https://discuss.huggingface.co/t/what-are-you-using-the-mini-bioasq-dataset-for/89042?u=tillwenke) you can share what you used the dataset for.
+
+Derives from http://participants-area.bioasq.org/Tasks/11b/trainingDataset/ we generated our own subset using `generate.py`.

data/passages.parquet/part.0.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93afd23506e6ab65451a43a411fd6481b6378a00fef7983b44cc8c3ddbb37c84
+size 24470604

data/test.parquet/part.0.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12679e03615d16b423b5554f8b2a6eb334f4cad89e62d202da5cc43cb9aeafb0
+size 1290026

generate.py

@@ -0,0 +1,93 @@
+import json
+
+import pandas as pd
+from Bio import Entrez
+from retry import retry
+from tqdm import tqdm
+import dask.dataframe as dd
+
+# provided your NIH credentials
+# read from .json file
+with open("credentials.json") as f:
+    credentials = json.load(f)
+    Entrez.email = credentials["email"]
+    Entrez.api_key = credentials["api_key"]
+
+
+# change output file names here if necessary
+RAW_EVALUATION_DATASET = "./raw_data/training11b.json"
+PATH_TO_PASSAGE_DATASET = "./data/passages.parquet"
+PATH_TO_EVALUATION_DATASET = "./data/test.parquet"
+
+# only use questions that have at most MAX_PASSAGES passages to control the size of the dataset
+# set to None to use all questions
+MAX_PASSAGES = None
+
+
+@retry()
+def get_abstract(passage_id):
+    with Entrez.efetch(
+        db="pubmed", id=passage_id, rettype="abstract", retmode="text"
+    ) as response:
+        # get only the abstract - no metadata
+        r = response.read()
+        r = r.split("\n\n")
+        abstract = max(r, key=len)
+        return abstract
+
+
+if __name__ == "__main__":
+    # load the training data containing the questions, answers and the ids of relevant passages
+    # but lacks the actual passages
+    with open(RAW_EVALUATION_DATASET) as f:
+        eval_data = json.load(f)["questions"]
+
+    eval_df = pd.DataFrame(eval_data, columns=["body", "documents", "ideal_answer"])
+    eval_df = eval_df.rename(
+        columns={
+            "body": "question",
+            "documents": "relevant_passage_ids",
+            "ideal_answer": "answer",
+        }
+    )
+    eval_df.answer = eval_df.answer.apply(lambda x: x[0])
+    # get abstract id from url
+    eval_df.relevant_passage_ids = eval_df.relevant_passage_ids.apply(
+        lambda x: [int(url.split("/")[-1]) for url in x]
+    )
+    if MAX_PASSAGES:
+        eval_df["passage_count"] = eval_df.relevant_passage_ids.apply(lambda x: len(x))
+        eval_df = eval_df.drop(columns=["passage_count"])
+
+    # remove duplicate passage ids
+    eval_df.relevant_passage_ids = eval_df.relevant_passage_ids.apply(lambda x: set(x))
+    eval_df.relevant_passage_ids = eval_df.relevant_passage_ids.apply(lambda x: list(x))
+
+    # get all passage ids that are relevant
+    passage_ids = set().union(*eval_df.relevant_passage_ids)
+    passage_ids = list(passage_ids)
+    passages = pd.DataFrame(index=passage_ids)
+
+    for i, passage_id in enumerate(tqdm(passages.index)):
+        passages.loc[passage_id, "passage"] = get_abstract(passage_id)
+
+        # intermediate save
+        if i % 1000 == 0:
+            passages.index.name = "id"
+            dd.from_pandas(passages, npartitions=1).to_parquet(PATH_TO_PASSAGE_DATASET)
+
+
+    # filter out the passages whos pmids (pubmed ids) where not available
+    unavailable_passages = passages[passages["passage"] == "1. "]
+    passages = passages[passages["passage"] != "1. "]
+    passages.index.name = "id"
+    dd.from_pandas(passages, npartitions=1).to_parquet(PATH_TO_PASSAGE_DATASET)
+
+    # remove passages from evaluation dataset whose abstract could not be retrieved from pubmed website
+    unavailable_ids = unavailable_passages.index.tolist()
+    eval_df["relevant_passage_ids"] = eval_df["relevant_passage_ids"].apply(
+        lambda x: [i for i in x if i not in unavailable_ids]
+    )
+    eval_df.index.name = "id"
+    eval_df = eval_df[["question", "answer", "relevant_passage_ids"]]
+    dd.from_pandas(eval_df, npartitions=1).to_parquet(PATH_TO_EVALUATION_DATASET)

raw_data/training11b.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6df656862ca860efc355c7805d07ddca700d64ecc3785c519a49afccaaeeac98
+size 37639648

requirements.txt

@@ -0,0 +1,50 @@
+asttokens==2.4.1
+backcall==0.2.0
+biopython==1.81
+click==8.1.7
+cloudpickle==3.0.0
+comm==0.1.4
+dask==2023.10.1
+debugpy==1.8.0
+decorator==5.1.1
+exceptiongroup==1.1.3
+executing==2.0.0
+fsspec==2023.10.0
+importlib-metadata==6.8.0
+ipykernel==6.26.0
+ipython==8.16.1
+jedi==0.19.1
+jupyter_client==8.5.0
+jupyter_core==5.4.0
+locket==1.0.0
+matplotlib-inline==0.1.6
+nest-asyncio==1.5.8
+numpy==1.26.1
+packaging==23.2
+pandas==2.1.2
+parso==0.8.3
+partd==1.4.1
+pexpect==4.8.0
+pickleshare==0.7.5
+platformdirs==3.11.0
+prompt-toolkit==3.0.39
+psutil==5.9.6
+ptyprocess==0.7.0
+pure-eval==0.2.2
+py==1.11.0
+pyarrow==13.0.0
+Pygments==2.16.1
+python-dateutil==2.8.2
+pytz==2023.3.post1
+PyYAML==6.0.1
+pyzmq==25.1.1
+retry==0.9.2
+six==1.16.0
+stack-data==0.6.3
+toolz==0.12.0
+tornado==6.3.3
+tqdm==4.66.1
+traitlets==5.12.0
+tzdata==2023.3
+wcwidth==0.2.8
+zipp==3.17.0