Add Chemdner dataset loader

[qanastek]

Please name your PR after the issue it closes. You can use the following line: "Closes #ISSUE-NUMBER" where you replace the ISSUE-NUMBER with the one corresponding to your dataset.

If the following information is NOT present in the issue, please populate:

• Name: name of the dataset
• Description: short description of the dataset (or link to social media or blog post)
• Paper: link to the dataset paper if available
• Data: link to the online home of the dataset

Checkbox

• Confirm that this PR is linked to the dataset issue.
• Create the dataloader script biodatasets/my_dataset/my_dataset.py (please use only lowercase and underscore for dataset naming).
• Provide values for the _CITATION, _DATASETNAME, _DESCRIPTION, _HOMEPAGE, _LICENSE, _URLs, _SUPPORTED_TASKS, _SOURCE_VERSION, and _BIGBIO_VERSION variables.
• Implement _info(), _split_generators() and _generate_examples() in dataloader script.
• Make sure that the BUILDER_CONFIGS class attribute is a list with at least one BigBioConfig for the source schema and one for a bigbio schema.
• Confirm dataloader script works with datasets.load_dataset function.
• Confirm that your dataloader script passes the test suite run with python -m tests.test_bigbio biodatasets/my_dataset/my_dataset.py.

[qanastek]

Issue #28

[qanastek]

Commit 526e7d9

[sg-wbi]

Thank you for implementing the changes! Everything looks fine. You made me realize that we could use the "MeSH_Indexing_Chemical " as labels for a text_classification schema. I asked internally about this and get back to you asap.

[qanastek]

Commit 6251db8

[galtay]

@sg-wbi thanks for self assigning this one! I'll jump to some of the other PRs

[galtay]

just a quick note, we have some extra files in this PR besides just the chemdner loader

[qanastek]

Fixed! I remove other files in the Commit 7796654.

[sg-wbi]

Thanks galtey! There is an updated on the schema for text_classification which allows for mulitple labels. See the "#general" channel in Discord.

@qanastek Could you please add this view to the dataset? You can use this PR #348 as an example.

Then we can close this!

[qanastek]

During the tests of the Text Classification task, I crashed during the datasets.load_dataset() function. Have you any ideas ? I use your code base.

The error :

======================================================================
ERROR: runTest (__main__.TestDataLoader)
Run all tests that check:
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/mnt/d/Projects/LIA/huggingface/BigScience BioMedical/pull_resquests/Final_ChemdNER/biomedical/tests/test_bigbio.py", line 164, in setUp
    self.datasets_bigbio[schema] = datasets.load_dataset(
  File "/home/home/miniconda3/envs/bigscience-biomedical/lib/python3.10/site-packages/datasets/load.py", line 1702, in load_dataset
    builder_instance.download_and_prepare(
  File "/home/home/miniconda3/envs/bigscience-biomedical/lib/python3.10/site-packages/datasets/builder.py", line 594, in download_and_prepare
    self._download_and_prepare(
  File "/home/home/miniconda3/envs/bigscience-biomedical/lib/python3.10/site-packages/datasets/builder.py", line 683, in _download_and_prepare
    self._prepare_split(split_generator, **prepare_split_kwargs)
  File "/home/home/miniconda3/envs/bigscience-biomedical/lib/python3.10/site-packages/datasets/builder.py", line 1080, in _prepare_split
    example = self.info.features.encode_example(record)
  File "/home/home/miniconda3/envs/bigscience-biomedical/lib/python3.10/site-packages/datasets/features/features.py", line 1287, in encode_example
    return encode_nested_example(self, example)
  File "/home/home/miniconda3/envs/bigscience-biomedical/lib/python3.10/site-packages/datasets/features/features.py", line 971, in encode_nested_example
    return {
  File "/home/home/miniconda3/envs/bigscience-biomedical/lib/python3.10/site-packages/datasets/features/features.py", line 971, in <dictcomp>
    return {
  File "/home/home/miniconda3/envs/bigscience-biomedical/lib/python3.10/site-packages/datasets/utils/py_utils.py", line 152, in zip_dict
    yield key, tuple(d[key] for d in dicts)
  File "/home/home/miniconda3/envs/bigscience-biomedical/lib/python3.10/site-packages/datasets/utils/py_utils.py", line 152, in <genexpr>
    yield key, tuple(d[key] for d in dicts)
KeyError: 'label'

----------------------------------------------------------------------
Ran 1 test in 15.950s

FAILED (errors=1)

My code :

    def _get_textcls_example(self, d: bioc.BioCDocument) -> Dict:

        example = {"document_id": d.id, "text": [], "labels": []}

        for p in d.passages:

            example["text"].append(p.text)

            for a in p.annotations:

                if a.infons.get("type") == "MeSH_Indexing_Chemical":

                    example["labels"].append(a.infons.get("identifier"))

        example["text"] = " ".join(example["text"])

        return example

The datasets.load_dataset is searching for a label field, except that we are giving it a labels one.

Have you passed the tests with the loading script of nlmchem ?

[sg-wbi]

The test for nlmchem passes. Are you working on the latest version of the repo?

git pull upstream master

[qanastek]

Thanks, It works.

[qanastek]

Commit 540901f

[sg-wbi]

Thank you very much for implementing the additional view! One last thing and then it should be good to go, for real this time. If I run the tests I get these warnings:

WARNING:__main__:Example:11379 - entity:11371  text:`None` != text_by_offset:``
Example:17134 - entity:17129  text:`None` != text_by_offset:``
Example:21504 - entity:21503  text:`None` != text_by_offset:``
Example:23227 - entity:23213  text:`None` != text_by_offset:``

It seems that somehow some entities without text get into the dataset. Could you please check again your script and fix this?
Thank you!

[qanastek]

Nice job! I haven't seemed them since the output is very compact.

It was due to None presents in the data :

• {'id': '21503', 'type': 'Chemical', 'text': [None], 'offsets': [[1607, 1607]], 'normalized': []}
• {'id': '11371', 'type': 'Chemical', 'text': [None], 'offsets': [[179, 179]], 'normalized': []}
• {'id': '17129', 'type': 'Chemical', 'text': [None], 'offsets': [[55, 55]], 'normalized': []}

I fixed it using a condition :

biomedical/biodatasets/chemdner/chemdner.py

Lines 293 to 294 in ac8fb5e

	if a.text == None or a.text == "":
	continue

I hope it's ok!

Commit ac8fb5e

[sg-wbi]

@qanastek

[qanastek]

Like this ?

if self.config.schema == "bigbio_kb" and a_type == "MeSH_Indexing_Chemical":
    continue

if (a.text == None or a.text == "") and self.config.schema == "bigbio_kb":
    continue

Or like this ?

if (self.config.schema == "bigbio_kb" and a_type == "MeSH_Indexing_Chemical") or ( (a.text == None or a.text == "") and self.config.schema == "bigbio_kb"):
    continue

[sg-wbi]

The first one is easier to read

[qanastek]

Commit 9cf9100

[sg-wbi]

Great! Thank you for bearing all the changes, especially w/ the schema change. Thanks again for your contribution!

[qanastek]

@sg-wbi Can you switch the issue (#28) status to Done to keep track of the advancement ?

[sg-wbi]

Sorry! I got confused and closed it! How to I change the status of #28? Can I simply close it?

[qanastek]

I think so, like Gabriel do on the Issue #53

[sg-wbi]

Now?

[qanastek]

Perfect! 🎉