← home
Github: datasets/medline.py

ir_datasets: Medline

Index
  1. medline
  2. medline/2004
  3. medline/2004/trec-genomics-2004
  4. medline/2004/trec-genomics-2005
  5. medline/2017
  6. medline/2017/trec-pm-2017
  7. medline/2017/trec-pm-2018

"medline"

Medical articles from Medline. This collection was used by TREC Genomics 2004-05 (2004 version of dataset) and by TREC Precision Medicine 2017-18 (2017 version).

"medline/2004"

3M Medline articles including titles and abstracts, used for the TREC 2004-05 Genomics track.

docs
3.7M docs

Language: en

Document type:
MedlineDoc: (namedtuple)
  1. doc_id: str
  2. title: str
  3. abstract: str

Examples:

Python API 
import ir_datasets
dataset = ir_datasets.load("medline/2004")
for doc in dataset.docs_iter():
    doc # namedtuple<doc_id, title, abstract>

You can find more details about the Python API here.

CLI 
ir_datasets export medline/2004 docs
[doc_id]    [title]    [abstract]
...

You can find more details about the CLI here.

PyTerrier 
import pyterrier as pt
pt.init()
dataset = pt.get_dataset('irds:medline/2004')
# Index medline/2004
indexer = pt.IterDictIndexer('./indices/medline_2004')
index_ref = indexer.index(dataset.get_corpus_iter(), fields=['title', 'abstract'])

You can find more details about PyTerrier indexing here.

Metadata 
{
  "docs": {
    "count": 3672808,
    "fields": {
      "doc_id": {
        "max_len": 8,
        "common_prefix": ""
      }
    }
  }
}

"medline/2004/trec-genomics-2004"

The TREC Genomics Track 2004 benchmark. Contains 50 queries with article-level relevance judgments.

queries
50 queries

Language: en

Query type:
TrecGenomicsQuery: (namedtuple)
  1. query_id: str
  2. title: str
  3. need: str
  4. context: str

Examples:

Python API 
import ir_datasets
dataset = ir_datasets.load("medline/2004/trec-genomics-2004")
for query in dataset.queries_iter():
    query # namedtuple<query_id, title, need, context>

You can find more details about the Python API here.

CLI 
ir_datasets export medline/2004/trec-genomics-2004 queries
[query_id]    [title]    [need]    [context]
...

You can find more details about the CLI here.

PyTerrier 
import pyterrier as pt
pt.init()
dataset = pt.get_dataset('irds:medline/2004/trec-genomics-2004')
index_ref = pt.IndexRef.of('./indices/medline_2004') # assumes you have already built an index
pipeline = pt.BatchRetrieve(index_ref, wmodel='BM25')
# (optionally other pipeline components)
pipeline(dataset.get_topics('title'))

You can find more details about PyTerrier retrieval here.

docs
3.7M docs

Inherits docs from medline/2004

Language: en

Document type:
MedlineDoc: (namedtuple)
  1. doc_id: str
  2. title: str
  3. abstract: str

Examples:

Python API 
import ir_datasets
dataset = ir_datasets.load("medline/2004/trec-genomics-2004")
for doc in dataset.docs_iter():
    doc # namedtuple<doc_id, title, abstract>

You can find more details about the Python API here.

CLI 
ir_datasets export medline/2004/trec-genomics-2004 docs
[doc_id]    [title]    [abstract]
...

You can find more details about the CLI here.

PyTerrier 
import pyterrier as pt
pt.init()
dataset = pt.get_dataset('irds:medline/2004/trec-genomics-2004')
# Index medline/2004
indexer = pt.IterDictIndexer('./indices/medline_2004')
index_ref = indexer.index(dataset.get_corpus_iter(), fields=['title', 'abstract'])

You can find more details about PyTerrier indexing here.

qrels
8.3K qrels
Query relevance judgment type:
TrecQrel: (namedtuple)
  1. query_id: str
  2. doc_id: str
  3. relevance: int
  4. iteration: str

Relevance levels

Rel.DefinitionCount%
0not relevant0 0.0%
1possibly relevant4.6K56.0%
2definitely relevant3.6K44.0%

Examples:

Python API 
import ir_datasets
dataset = ir_datasets.load("medline/2004/trec-genomics-2004")
for qrel in dataset.qrels_iter():
    qrel # namedtuple<query_id, doc_id, relevance, iteration>

You can find more details about the Python API here.

CLI 
ir_datasets export medline/2004/trec-genomics-2004 qrels --format tsv
[query_id]    [doc_id]    [relevance]    [iteration]
...

You can find more details about the CLI here.

PyTerrier 
import pyterrier as pt
from pyterrier.measures import *
pt.init()
dataset = pt.get_dataset('irds:medline/2004/trec-genomics-2004')
index_ref = pt.IndexRef.of('./indices/medline_2004') # assumes you have already built an index
pipeline = pt.BatchRetrieve(index_ref, wmodel='BM25')
# (optionally other pipeline components)
pt.Experiment(
    [pipeline],
    dataset.get_topics('title'),
    dataset.get_qrels(),
    [MAP, nDCG@20]
)

You can find more details about PyTerrier experiments here.

Citation

ir_datasets.bib:

\cite{Hersh2004TrecGenomics}

Bibtex:

@inproceedings{Hersh2004TrecGenomics, title={TREC 2004 Genomics Track Overview}, author={William R. Hersh and Ravi Teja Bhuptiraju and Laura Ross and Phoebe Johnson and Aaron M. Cohen and Dale F. Kraemer}, booktitle={TREC}, year={2004} }
Metadata 
{
  "docs": {
    "count": 3672808,
    "fields": {
      "doc_id": {
        "max_len": 8,
        "common_prefix": ""
      }
    }
  },
  "queries": {
    "count": 50
  },
  "qrels": {
    "count": 8268,
    "fields": {
      "relevance": {
        "counts_by_value": {
          "2": 3639,
          "1": 4629
        }
      }
    }
  }
}

"medline/2004/trec-genomics-2005"

The TREC Genomics Track 2005 benchmark. Contains 36 queries with passage-level relevance judgments.

queries
50 queries

Language: en

Query type:
GenericQuery: (namedtuple)
  1. query_id: str
  2. text: str

Examples:

Python API 
import ir_datasets
dataset = ir_datasets.load("medline/2004/trec-genomics-2005")
for query in dataset.queries_iter():
    query # namedtuple<query_id, text>

You can find more details about the Python API here.

CLI 
ir_datasets export medline/2004/trec-genomics-2005 queries
[query_id]    [text]
...

You can find more details about the CLI here.

PyTerrier 
import pyterrier as pt
pt.init()
dataset = pt.get_dataset('irds:medline/2004/trec-genomics-2005')
index_ref = pt.IndexRef.of('./indices/medline_2004') # assumes you have already built an index
pipeline = pt.BatchRetrieve(index_ref, wmodel='BM25')
# (optionally other pipeline components)
pipeline(dataset.get_topics())

You can find more details about PyTerrier retrieval here.

docs
3.7M docs

Inherits docs from medline/2004

Language: en

Document type:
MedlineDoc: (namedtuple)
  1. doc_id: str
  2. title: str
  3. abstract: str

Examples:

Python API 
import ir_datasets
dataset = ir_datasets.load("medline/2004/trec-genomics-2005")
for doc in dataset.docs_iter():
    doc # namedtuple<doc_id, title, abstract>

You can find more details about the Python API here.

CLI 
ir_datasets export medline/2004/trec-genomics-2005 docs
[doc_id]    [title]    [abstract]
...

You can find more details about the CLI here.

PyTerrier 
import pyterrier as pt
pt.init()
dataset = pt.get_dataset('irds:medline/2004/trec-genomics-2005')
# Index medline/2004
indexer = pt.IterDictIndexer('./indices/medline_2004')
index_ref = indexer.index(dataset.get_corpus_iter(), fields=['title', 'abstract'])

You can find more details about PyTerrier indexing here.

qrels
40K qrels
Query relevance judgment type:
TrecQrel: (namedtuple)
  1. query_id: str
  2. doc_id: str
  3. relevance: int
  4. iteration: str

Relevance levels

Rel.DefinitionCount%
0not relevant35K88.5%
1possibly relevant2.1K5.2%
2definitely relevant2.5K6.3%

Examples:

Python API 
import ir_datasets
dataset = ir_datasets.load("medline/2004/trec-genomics-2005")
for qrel in dataset.qrels_iter():
    qrel # namedtuple<query_id, doc_id, relevance, iteration>

You can find more details about the Python API here.

CLI 
ir_datasets export medline/2004/trec-genomics-2005 qrels --format tsv
[query_id]    [doc_id]    [relevance]    [iteration]
...

You can find more details about the CLI here.

PyTerrier 
import pyterrier as pt
from pyterrier.measures import *
pt.init()
dataset = pt.get_dataset('irds:medline/2004/trec-genomics-2005')
index_ref = pt.IndexRef.of('./indices/medline_2004') # assumes you have already built an index
pipeline = pt.BatchRetrieve(index_ref, wmodel='BM25')
# (optionally other pipeline components)
pt.Experiment(
    [pipeline],
    dataset.get_topics(),
    dataset.get_qrels(),
    [MAP, nDCG@20]
)

You can find more details about PyTerrier experiments here.

Citation

ir_datasets.bib:

\cite{Hersh2005TrecGenomics}

Bibtex:

@inproceedings{Hersh2005TrecGenomics, title={TREC 2005 Genomics Track Overview}, author={William Hersh and Aaron Cohen and Jianji Yang and Ravi Teja Bhupatiraju and Phoebe Roberts and Marti Hearst}, booktitle={TREC}, year={2007} }
Metadata 
{
  "docs": {
    "count": 3672808,
    "fields": {
      "doc_id": {
        "max_len": 8,
        "common_prefix": ""
      }
    }
  },
  "queries": {
    "count": 50
  },
  "qrels": {
    "count": 39958,
    "fields": {
      "relevance": {
        "counts_by_value": {
          "0": 35374,
          "2": 2525,
          "1": 2059
        }
      }
    }
  }
}

"medline/2017"

26M Medline and AACR/ASCO Proceedings articles including titles and abstracts. This collection is used for the TREC 2017-18 TREC Precision Medicine track.

docs
27M docs

Language: en

Document type:
MedlineDoc: (namedtuple)
  1. doc_id: str
  2. title: str
  3. abstract: str

Examples:

Python API 
import ir_datasets
dataset = ir_datasets.load("medline/2017")
for doc in dataset.docs_iter():
    doc # namedtuple<doc_id, title, abstract>

You can find more details about the Python API here.

CLI 
ir_datasets export medline/2017 docs
[doc_id]    [title]    [abstract]
...

You can find more details about the CLI here.

PyTerrier 
import pyterrier as pt
pt.init()
dataset = pt.get_dataset('irds:medline/2017')
# Index medline/2017
indexer = pt.IterDictIndexer('./indices/medline_2017')
index_ref = indexer.index(dataset.get_corpus_iter(), fields=['title', 'abstract'])

You can find more details about PyTerrier indexing here.

Metadata 
{
  "docs": {
    "count": 26740025,
    "fields": {
      "doc_id": {
        "max_len": 15,
        "common_prefix": ""
      }
    }
  }
}

"medline/2017/trec-pm-2017"

The TREC Precision Medicine (PM) Track 2017 benchmark. Contains 30 queries containing disease, gene, and target demographic information.

queries
30 queries

Language: en

Query type:
TrecPm2017Query: (namedtuple)
  1. query_id: str
  2. disease: str
  3. gene: str
  4. demographic: str
  5. other: str

Examples:

Python API 
import ir_datasets
dataset = ir_datasets.load("medline/2017/trec-pm-2017")
for query in dataset.queries_iter():
    query # namedtuple<query_id, disease, gene, demographic, other>

You can find more details about the Python API here.

CLI 
ir_datasets export medline/2017/trec-pm-2017 queries
[query_id]    [disease]    [gene]    [demographic]    [other]
...

You can find more details about the CLI here.

PyTerrier 
import pyterrier as pt
pt.init()
dataset = pt.get_dataset('irds:medline/2017/trec-pm-2017')
index_ref = pt.IndexRef.of('./indices/medline_2017') # assumes you have already built an index
pipeline = pt.BatchRetrieve(index_ref, wmodel='BM25')
# (optionally other pipeline components)
pipeline(dataset.get_topics('disease'))

You can find more details about PyTerrier retrieval here.

docs
27M docs

Inherits docs from medline/2017

Language: en

Document type:
MedlineDoc: (namedtuple)
  1. doc_id: str
  2. title: str
  3. abstract: str

Examples:

Python API 
import ir_datasets
dataset = ir_datasets.load("medline/2017/trec-pm-2017")
for doc in dataset.docs_iter():
    doc # namedtuple<doc_id, title, abstract>

You can find more details about the Python API here.

CLI 
ir_datasets export medline/2017/trec-pm-2017 docs
[doc_id]    [title]    [abstract]
...

You can find more details about the CLI here.

PyTerrier 
import pyterrier as pt
pt.init()
dataset = pt.get_dataset('irds:medline/2017/trec-pm-2017')
# Index medline/2017
indexer = pt.IterDictIndexer('./indices/medline_2017')
index_ref = indexer.index(dataset.get_corpus_iter(), fields=['title', 'abstract'])

You can find more details about PyTerrier indexing here.

qrels
23K qrels
Query relevance judgment type:
TrecQrel: (namedtuple)
  1. query_id: str
  2. doc_id: str
  3. relevance: int
  4. iteration: str

Relevance levels

Rel.DefinitionCount%
0not relevant19K82.9%
1possibly relevant1.9K8.2%
2definitely relevant2.0K8.9%

Examples:

Python API 
import ir_datasets
dataset = ir_datasets.load("medline/2017/trec-pm-2017")
for qrel in dataset.qrels_iter():
    qrel # namedtuple<query_id, doc_id, relevance, iteration>

You can find more details about the Python API here.

CLI 
ir_datasets export medline/2017/trec-pm-2017 qrels --format tsv
[query_id]    [doc_id]    [relevance]    [iteration]
...

You can find more details about the CLI here.

PyTerrier 
import pyterrier as pt
from pyterrier.measures import *
pt.init()
dataset = pt.get_dataset('irds:medline/2017/trec-pm-2017')
index_ref = pt.IndexRef.of('./indices/medline_2017') # assumes you have already built an index
pipeline = pt.BatchRetrieve(index_ref, wmodel='BM25')
# (optionally other pipeline components)
pt.Experiment(
    [pipeline],
    dataset.get_topics('disease'),
    dataset.get_qrels(),
    [MAP, nDCG@20]
)

You can find more details about PyTerrier experiments here.

Citation

ir_datasets.bib:

\cite{Roberts2017TrecPm}

Bibtex:

@inproceedings{Roberts2017TrecPm, title={Overview of the TREC 2017 Precision Medicine Track}, author={Kirk Roberts and Dina Demner-Fushman and Ellen M. Voorhees and William R. Hersh and Steven Bedrick and Alexander J. Lazar and Shubham Pant}, booktitle={TREC}, year={2017} }
Metadata 
{
  "docs": {
    "count": 26740025,
    "fields": {
      "doc_id": {
        "max_len": 15,
        "common_prefix": ""
      }
    }
  },
  "queries": {
    "count": 30
  },
  "qrels": {
    "count": 22642,
    "fields": {
      "relevance": {
        "counts_by_value": {
          "0": 18767,
          "1": 1853,
          "2": 2022
        }
      }
    }
  }
}

"medline/2017/trec-pm-2018"

The TREC Precision Medicine (PM) Track 2018 benchmark. Contains 50 queries containing disease, gene, and target demographic information.

queries
50 queries

Language: en

Query type:
TrecPmQuery: (namedtuple)
  1. query_id: str
  2. disease: str
  3. gene: str
  4. demographic: str

Examples:

Python API 
import ir_datasets
dataset = ir_datasets.load("medline/2017/trec-pm-2018")
for query in dataset.queries_iter():
    query # namedtuple<query_id, disease, gene, demographic>

You can find more details about the Python API here.

CLI 
ir_datasets export medline/2017/trec-pm-2018 queries
[query_id]    [disease]    [gene]    [demographic]
...

You can find more details about the CLI here.

PyTerrier 
import pyterrier as pt
pt.init()
dataset = pt.get_dataset('irds:medline/2017/trec-pm-2018')
index_ref = pt.IndexRef.of('./indices/medline_2017') # assumes you have already built an index
pipeline = pt.BatchRetrieve(index_ref, wmodel='BM25')
# (optionally other pipeline components)
pipeline(dataset.get_topics('disease'))

You can find more details about PyTerrier retrieval here.

docs
27M docs

Inherits docs from medline/2017

Language: en

Document type:
MedlineDoc: (namedtuple)
  1. doc_id: str
  2. title: str
  3. abstract: str

Examples:

Python API 
import ir_datasets
dataset = ir_datasets.load("medline/2017/trec-pm-2018")
for doc in dataset.docs_iter():
    doc # namedtuple<doc_id, title, abstract>

You can find more details about the Python API here.

CLI 
ir_datasets export medline/2017/trec-pm-2018 docs
[doc_id]    [title]    [abstract]
...

You can find more details about the CLI here.

PyTerrier 
import pyterrier as pt
pt.init()
dataset = pt.get_dataset('irds:medline/2017/trec-pm-2018')
# Index medline/2017
indexer = pt.IterDictIndexer('./indices/medline_2017')
index_ref = indexer.index(dataset.get_corpus_iter(), fields=['title', 'abstract'])

You can find more details about PyTerrier indexing here.

qrels
22K qrels
Query relevance judgment type:
TrecQrel: (namedtuple)
  1. query_id: str
  2. doc_id: str
  3. relevance: int
  4. iteration: str

Relevance levels

Rel.DefinitionCount%
0not relevant17K75.1%
1possibly relevant2.1K9.6%
2definitely relevant3.4K15.3%

Examples:

Python API 
import ir_datasets
dataset = ir_datasets.load("medline/2017/trec-pm-2018")
for qrel in dataset.qrels_iter():
    qrel # namedtuple<query_id, doc_id, relevance, iteration>

You can find more details about the Python API here.

CLI 
ir_datasets export medline/2017/trec-pm-2018 qrels --format tsv
[query_id]    [doc_id]    [relevance]    [iteration]
...

You can find more details about the CLI here.

PyTerrier 
import pyterrier as pt
from pyterrier.measures import *
pt.init()
dataset = pt.get_dataset('irds:medline/2017/trec-pm-2018')
index_ref = pt.IndexRef.of('./indices/medline_2017') # assumes you have already built an index
pipeline = pt.BatchRetrieve(index_ref, wmodel='BM25')
# (optionally other pipeline components)
pt.Experiment(
    [pipeline],
    dataset.get_topics('disease'),
    dataset.get_qrels(),
    [MAP, nDCG@20]
)

You can find more details about PyTerrier experiments here.

Citation

ir_datasets.bib:

\cite{Roberts2018TrecPm}

Bibtex:

@inproceedings{Roberts2018TrecPm, title={Overview of the TREC 2018 Precision Medicine Track}, author={Kirk Roberts and Dina Demner-Fushman and Ellen M. Voorhees and William R. Hersh and Steven Bedrick and Alexander J. Lazar}, booktitle={TREC}, year={2018} }
Metadata 
{
  "docs": {
    "count": 26740025,
    "fields": {
      "doc_id": {
        "max_len": 15,
        "common_prefix": ""
      }
    }
  },
  "queries": {
    "count": 50
  },
  "qrels": {
    "count": 22429,
    "fields": {
      "relevance": {
        "counts_by_value": {
          "0": 16841,
          "2": 3442,
          "1": 2146
        }
      }
    }
  }
}