Datasets:

VelkroLM
/

hausa-stopwords-corpus

Languages:

Hausa

Tags:

hausa

stopwords

nlp

low-resource-language

License:

Dataset card Files Files and versions

xet

Community

Dataset Viewer

The dataset viewer is not available for this subset.

Cannot get the split names for the config 'default' of the dataset.

Exception:    SplitsNotFoundError
Message:      The split names could not be parsed from the dataset config.
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 290, in _generate_tables
                  pa_table = paj.read_json(
                      io.BytesIO(batch), read_options=paj.ReadOptions(block_size=block_size)
                  )
                File "pyarrow/_json.pyx", line 342, in pyarrow._json.read_json
                File "pyarrow/error.pxi", line 155, in pyarrow.lib.pyarrow_internal_check_status
                  return check_status(status)
                File "pyarrow/error.pxi", line 92, in pyarrow.lib.check_status
                  raise convert_status(status)
              pyarrow.lib.ArrowInvalid: JSON parse error: Column() changed from object to string in row 0
              
              During handling of the above exception, another exception occurred:
              
              Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
                  for split_generator in builder._split_generators(
                                         ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 101, in _split_generators
                  pa_table = next(iter(self._generate_tables(**splits[0].gen_kwargs, allow_full_read=False)))[1]
                             ~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 304, in _generate_tables
                  batch = json_encode_fields_in_json_lines(original_batch, json_field_paths)
                File "/usr/local/lib/python3.14/site-packages/datasets/utils/json.py", line 111, in json_encode_fields_in_json_lines
                  examples = [ujson_loads(line) for line in original_batch.splitlines()]
                              ~~~~~~~~~~~^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/utils/json.py", line 20, in ujson_loads
                  return pd.io.json.ujson_loads(*args, **kwargs)
                         ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
              ValueError: Expected object or value
              
              The above exception was the direct cause of the following exception:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 66, in compute_split_names_from_streaming_response
                  for split in get_dataset_split_names(
                               ~~~~~~~~~~~~~~~~~~~~~~~^
                      path=dataset,
                      ^^^^^^^^^^^^^
                      config_name=config,
                      ^^^^^^^^^^^^^^^^^^^
                      token=hf_token,
                      ^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
                  info = get_dataset_config_info(
                      path,
                  ...<6 lines>...
                      **config_kwargs,
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
                  raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
              datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

Hausa Stopwords Corpus (VelkroLM)

Status: In progress -- full word scoring complete, AI-based English/name filtering and news-bias adjustment not yet applied.

What this is

A cross-source Hausa stopword candidate list, built by pooling multiple verified Hausa text sources into a single corpus and scoring every word by document frequency. This is a candidate list with scores, not a final curated stopword list -- the goal is to publish full scores so others can compare against their own data, correlate results, and build their own tools or libraries on top of it.

Sources used

All sources were manually verified before inclusion:

CC-100 Hausa -- web crawl text, direct from data.statmt.org/cc-100/ha.txt.xz
Wikipedia (Hausa) -- wikimedia/wikipedia, config 20231101.ha
Leipzig Corpora Collection (Hausa Community 2017) -- blogs/forums/web text
QA/instruction pairs (rufatronics/mt5_hausa) -- clean, human-reviewed instruction-style Hausa prose

All four sources are pooled into one merged corpus -- there is no per-source/genre weighting in the scoring. Every word gets a single score based on how often it appears across the whole combined corpus.

Repository structure


raw_scores/
full_word_scores.json      <- every scored word, full metrics, no cutoff
batches/
batch_0001.json ...        <- the full word list split into 3,000-word
batches, ranked by score (for browsing/
downstream processing in smaller chunks)
top1000/
top_1000_candidates.json/.csv/.txt   <- quick-view top 1000 by score
ai_scan_results/
(placeholder -- see folder README)   <- Gemini-based English-word/name
filter results go here once run
news_bias_processed/
(placeholder -- see folder README)   <- news-bias-adjusted scores for
the full word list go here once run
final_stopwords/
(placeholder -- see folder README)   <- final human-reviewed stopword
count/list goes here once finalized
docs/
METHODOLOGY.md             <- full write-up of scoring approach and the
planned news-bias correction methodology

Score fields (in `raw_scores/full_word_scores.json` and `batches/`)

Each word has:

document_frequency -- how many lines (documents) in the merged corpus contain the word
document_frequency_ratio -- that count divided by total documents in the corpus
term_frequency -- total raw occurrences across the merged corpus
composite_score -- currently equal to document_frequency_ratio (the simplest, most direct measure of "how common is this word across everything")

Once the AI-based filter and news-bias adjustment are applied, additional fields (is_english_or_name, adjusted_score, news_bias_flag, bias_adjusted_score) will appear in ai_scan_results/ and news_bias_processed/ respectively -- see those folders' own README files and docs/METHODOLOGY.md for details on each.

Why publish full scores, not just a final list

The intent is for other people working on Hausa NLP to be able to compare this scoring against their own data/corpora and check for correlation, rather than just trusting a fixed, opaque stopword list. A library or tool may be built on top of this later.

License / usage

See individual source licenses (CC-100, Wikipedia, Leipzig Corpora, and the QA/instruction dataset each have their own terms) -- this repository aggregates scores derived from them, not the raw source text itself (beyond what's needed for scoring transparency).