Hugging Face's logo

Datasets:
kev-KOH
/
time-embed-bge-m3

Tasks:
Sentence Similarity
Feature Extraction
Languages:
Korean
Size:
10K<n<100K
License:
Dataset card Files
xet
 Community
Dataset Viewer
Duplicate
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
query: string
pos: list<item: string>
  child 0, item: string
neg: list<item: string>
  child 0, item: string
metadata: struct<benchmark_set: string, focus_bucket: string, negative_types: list<item: string>, negative_typ (... 306 chars omitted)
  child 0, benchmark_set: string
  child 1, focus_bucket: string
  child 2, negative_types: list<item: string>
      child 0, item: string
  child 3, negative_type_sequence: list<item: string>
      child 0, item: string
  child 4, calendar_granularity: string
  child 5, reference_date_case: string
  child 6, reference_date: timestamp[s]
  child 7, coincidence_label: string
  child 8, query_intent: string
  child 9, target_time_field: string
  child 10, target_period: string
  child 11, temporal_relation: string
  child 12, surface_family: string
  child 13, difficulty: string
  child 14, source: string
negative_type_counts: struct<calendar_vs_duration: int64, calendar_vs_window: int64, adjacent_calendar_period: int64, cale (... 60 chars omitted)
  child 0, calendar_vs_duration: int64
  child 1, calendar_vs_window: int64
  child 2, adjacent_calendar_period: int64
  child 3, calendar_vs_month_offset: int64
  child 4, calendar_vs_year_offset: int64
split_counts: struct<test: int64>
  child 0, test: int64
difficulty_counts: struct<hard_boundary: int64, conditional_hard: int64>
  child 0, hard_boundary: int64
  child 1, conditional_hard: int64
reference_date_case_counts: struct<week_tuesday_after_start: int64, week_year_crossing: int64, we
...
int64>
  child 0, calendar_week: int64
  child 1, calendar_month: int64
  child 2, calendar_year: int64
temporal_relation_counts: struct<calendar_range: int64>
  child 0, calendar_range: int64
benchmark_sets: list<item: string>
  child 0, item: string
seed: int64
rows_per_benchmark_set: int64
total_records: int64
focus_bucket_counts: struct<calendar_week_vs_duration: int64, calendar_week_adjacent_period: int64, calendar_week_vs_wind (... 273 chars omitted)
  child 0, calendar_week_vs_duration: int64
  child 1, calendar_week_adjacent_period: int64
  child 2, calendar_week_vs_window: int64
  child 3, calendar_month_vs_duration: int64
  child 4, calendar_month_adjacent_period: int64
  child 5, calendar_month_vs_month_offset: int64
  child 6, calendar_year_vs_duration: int64
  child 7, calendar_year_adjacent_period: int64
  child 8, calendar_year_vs_year_offset: int64
  child 9, conditional_coincidence_boundary: int64
calendar_granularity_counts: struct<week: int64, month: int64, year: int64>
  child 0, week: int64
  child 1, month: int64
  child 2, year: int64
coincidence_label_counts: struct<non_coincident: int64, not_applicable: int64, coincident: int64>
  child 0, non_coincident: int64
  child 1, not_applicable: int64
  child 2, coincident: int64
query_intent_counts: struct<find_uploaded_material: int64, find_recent_notice: int64, find_upcoming_lecture: int64>
  child 0, find_uploaded_material: int64
  child 1, find_recent_notice: int64
  child 2, find_upcoming_lecture: int64
to
{'total_records': Value('int64'), 'split_counts': {'test': Value('int64')}, 'rows_per_benchmark_set': Value('int64'), 'seed': Value('int64'), 'benchmark_set_counts': {'calendar_week_boundary_test': Value('int64'), 'calendar_month_boundary_test': Value('int64'), 'calendar_year_boundary_test': Value('int64'), 'conditional_coincidence_test': Value('int64')}, 'focus_bucket_counts': {'calendar_week_vs_duration': Value('int64'), 'calendar_week_adjacent_period': Value('int64'), 'calendar_week_vs_window': Value('int64'), 'calendar_month_vs_duration': Value('int64'), 'calendar_month_adjacent_period': Value('int64'), 'calendar_month_vs_month_offset': Value('int64'), 'calendar_year_vs_duration': Value('int64'), 'calendar_year_adjacent_period': Value('int64'), 'calendar_year_vs_year_offset': Value('int64'), 'conditional_coincidence_boundary': Value('int64')}, 'negative_type_counts': {'calendar_vs_duration': Value('int64'), 'calendar_vs_window': Value('int64'), 'adjacent_calendar_period': Value('int64'), 'calendar_vs_month_offset': Value('int64'), 'calendar_vs_year_offset': Value('int64')}, 'calendar_granularity_counts': {'week': Value('int64'), 'month': Value('int64'), 'year': Value('int64')}, 'reference_date_case_counts': {'week_tuesday_after_start': Value('int64'), 'week_year_crossing': Value('int64'), 'week_sunday_end': Value('int64'), 'month_first_day_30_day_prev': Value('int64'), 'month_last_day_31': Value('int64'), 'month_feb_leap': Value('int64'), 'year_start': Value('int64'), 'leap_day': Value('int64'), 'year_end': Value('int64'), 'week_monday_start': Value('int64'), 'month_second_day_31_day_month': Value('int64'), 'after_leap_day': Value('int64')}, 'coincidence_label_counts': {'non_coincident': Value('int64'), 'not_applicable': Value('int64'), 'coincident': Value('int64')}, 'query_intent_counts': {'find_uploaded_material': Value('int64'), 'find_recent_notice': Value('int64'), 'find_upcoming_lecture': Value('int64')}, 'temporal_relation_counts': {'calendar_range': Value('int64')}, 'surface_family_counts': {'calendar_week': Value('int64'), 'calendar_month': Value('int64'), 'calendar_year': Value('int64')}, 'difficulty_counts': {'hard_boundary': Value('int64'), 'conditional_hard': Value('int64')}, 'benchmark_sets': List(Value('string'))}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 99, in get_rows_or_raise
                  return get_rows(
                         ^^^^^^^^^
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                         ^^^^^^^^^^^^^^^^^^^^^
                File "/src/services/worker/src/worker/utils.py", line 77, in get_rows
                  rows_plus_one = list(itertools.islice(ds, rows_max_number + 1))
                                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2690, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2227, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2251, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 494, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 384, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/packaged_modules/json/json.py", line 299, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/packaged_modules/json/json.py", line 128, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2321, in table_cast
                  return cast_table_to_schema(table, schema)
                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2249, in cast_table_to_schema
                  raise CastError(
              datasets.table.CastError: Couldn't cast
              query: string
              pos: list<item: string>
                child 0, item: string
              neg: list<item: string>
                child 0, item: string
              metadata: struct<benchmark_set: string, focus_bucket: string, negative_types: list<item: string>, negative_typ (... 306 chars omitted)
                child 0, benchmark_set: string
                child 1, focus_bucket: string
                child 2, negative_types: list<item: string>
                    child 0, item: string
                child 3, negative_type_sequence: list<item: string>
                    child 0, item: string
                child 4, calendar_granularity: string
                child 5, reference_date_case: string
                child 6, reference_date: timestamp[s]
                child 7, coincidence_label: string
                child 8, query_intent: string
                child 9, target_time_field: string
                child 10, target_period: string
                child 11, temporal_relation: string
                child 12, surface_family: string
                child 13, difficulty: string
                child 14, source: string
              negative_type_counts: struct<calendar_vs_duration: int64, calendar_vs_window: int64, adjacent_calendar_period: int64, cale (... 60 chars omitted)
                child 0, calendar_vs_duration: int64
                child 1, calendar_vs_window: int64
                child 2, adjacent_calendar_period: int64
                child 3, calendar_vs_month_offset: int64
                child 4, calendar_vs_year_offset: int64
              split_counts: struct<test: int64>
                child 0, test: int64
              difficulty_counts: struct<hard_boundary: int64, conditional_hard: int64>
                child 0, hard_boundary: int64
                child 1, conditional_hard: int64
              reference_date_case_counts: struct<week_tuesday_after_start: int64, week_year_crossing: int64, we
              ...
              int64>
                child 0, calendar_week: int64
                child 1, calendar_month: int64
                child 2, calendar_year: int64
              temporal_relation_counts: struct<calendar_range: int64>
                child 0, calendar_range: int64
              benchmark_sets: list<item: string>
                child 0, item: string
              seed: int64
              rows_per_benchmark_set: int64
              total_records: int64
              focus_bucket_counts: struct<calendar_week_vs_duration: int64, calendar_week_adjacent_period: int64, calendar_week_vs_wind (... 273 chars omitted)
                child 0, calendar_week_vs_duration: int64
                child 1, calendar_week_adjacent_period: int64
                child 2, calendar_week_vs_window: int64
                child 3, calendar_month_vs_duration: int64
                child 4, calendar_month_adjacent_period: int64
                child 5, calendar_month_vs_month_offset: int64
                child 6, calendar_year_vs_duration: int64
                child 7, calendar_year_adjacent_period: int64
                child 8, calendar_year_vs_year_offset: int64
                child 9, conditional_coincidence_boundary: int64
              calendar_granularity_counts: struct<week: int64, month: int64, year: int64>
                child 0, week: int64
                child 1, month: int64
                child 2, year: int64
              coincidence_label_counts: struct<non_coincident: int64, not_applicable: int64, coincident: int64>
                child 0, non_coincident: int64
                child 1, not_applicable: int64
                child 2, coincident: int64
              query_intent_counts: struct<find_uploaded_material: int64, find_recent_notice: int64, find_upcoming_lecture: int64>
                child 0, find_uploaded_material: int64
                child 1, find_recent_notice: int64
                child 2, find_upcoming_lecture: int64
              to
              {'total_records': Value('int64'), 'split_counts': {'test': Value('int64')}, 'rows_per_benchmark_set': Value('int64'), 'seed': Value('int64'), 'benchmark_set_counts': {'calendar_week_boundary_test': Value('int64'), 'calendar_month_boundary_test': Value('int64'), 'calendar_year_boundary_test': Value('int64'), 'conditional_coincidence_test': Value('int64')}, 'focus_bucket_counts': {'calendar_week_vs_duration': Value('int64'), 'calendar_week_adjacent_period': Value('int64'), 'calendar_week_vs_window': Value('int64'), 'calendar_month_vs_duration': Value('int64'), 'calendar_month_adjacent_period': Value('int64'), 'calendar_month_vs_month_offset': Value('int64'), 'calendar_year_vs_duration': Value('int64'), 'calendar_year_adjacent_period': Value('int64'), 'calendar_year_vs_year_offset': Value('int64'), 'conditional_coincidence_boundary': Value('int64')}, 'negative_type_counts': {'calendar_vs_duration': Value('int64'), 'calendar_vs_window': Value('int64'), 'adjacent_calendar_period': Value('int64'), 'calendar_vs_month_offset': Value('int64'), 'calendar_vs_year_offset': Value('int64')}, 'calendar_granularity_counts': {'week': Value('int64'), 'month': Value('int64'), 'year': Value('int64')}, 'reference_date_case_counts': {'week_tuesday_after_start': Value('int64'), 'week_year_crossing': Value('int64'), 'week_sunday_end': Value('int64'), 'month_first_day_30_day_prev': Value('int64'), 'month_last_day_31': Value('int64'), 'month_feb_leap': Value('int64'), 'year_start': Value('int64'), 'leap_day': Value('int64'), 'year_end': Value('int64'), 'week_monday_start': Value('int64'), 'month_second_day_31_day_month': Value('int64'), 'after_leap_day': Value('int64')}, 'coincidence_label_counts': {'non_coincident': Value('int64'), 'not_applicable': Value('int64'), 'coincident': Value('int64')}, 'query_intent_counts': {'find_uploaded_material': Value('int64'), 'find_recent_notice': Value('int64'), 'find_upcoming_lecture': Value('int64')}, 'temporal_relation_counts': {'calendar_range': Value('int64')}, 'surface_family_counts': {'calendar_week': Value('int64'), 'calendar_month': Value('int64'), 'calendar_year': Value('int64')}, 'difficulty_counts': {'hard_boundary': Value('int64'), 'conditional_hard': Value('int64')}, 'benchmark_sets': List(Value('string'))}
              because column names don't match

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.

Korean Temporal Query Embedding Data for BGE-M3 - Curated Template Augmented

This dataset is a FlagEmbedding/BGE-M3 fine-tuning dataset for Korean LMS temporal queries.

The goal is to make semantically equivalent Korean relative time expressions close in embedding space while separating similar-looking but temporally different expressions.

This version uses agent-curated Korean query templates with {time_expression} and {semantic_keyword} placeholders. Deterministic code inserted normalized temporal surfaces, created labels, and validated pair/triplet records.

LLM-generated templates may be used as proposals, but they are not trusted as final training data unless they pass intent-specific deterministic validation. Deadline templates that create malformed strings such as duplicate 까지 suffixes are excluded.

Examples:

  • Positive: 일주일 전 = 7일 전 = 1주 전
  • Hard negative: 일주일 전 != 지난주
  • Hard negative: 7일 전 != 최근 7일

Format

Each JSONL row follows the FlagEmbedding retrieval training format:

{
  "query": "1주 전 공개된 Hospitality 자료 목록 보여줘",
  "pos": ["7일 전 공개된 Hospitality 자료 목록 보여줘"],
  "neg": ["최근 7일 공개된 Hospitality 자료 목록 보여줘"]
}

Files 

train.jsonl
dev.jsonl
test.jsonl
bge_m3_training_stats.json

Counts

Split Rows
train 16,464
dev 2,568
test 2,640
total 21,672

Source Artifacts

The source curated-template LMS temporal embedding artifacts were validated before conversion:

templates: 120
pairs: 43,200
triplets: 12,000
validation: ok

The BGE-M3 conversion uses only rows whose source validation status is validated. Additional quality checks exclude malformed generated strings such as 내일까지까지, 하루 안에까지, and 내일까지 내에.

Intended Use

This dataset is intended for fine-tuning BAAI/bge-m3 with FlagEmbedding.finetune.embedder.encoder_only.m3.

Exclusions

Fuzzy temporal expressions such as vague "recently" or "soon" style labels are excluded from the main training labels.
Downloads last month
62
Total file size:
356 MB