Update README.md

README.md

@@ -1,233 +1,23 @@
 ---
-dataset_info:
-- config_name: corpus
-  features:
-  - name: pid
-    dtype: int64
-  - name: passage
-    dtype: string
-  splits:
-  - name: train
-    num_bytes: 3493263838
-    num_examples: 8473865
-  download_size: 1769907689
-  dataset_size: 3493263838
-- config_name: queries
-  features:
-  - name: qid
-    dtype: int64
-  - name: query
-    dtype: string
-  splits:
-  - name: train
-    num_bytes: 43751310
-    num_examples: 798077
-  download_size: 28090102
-  dataset_size: 43751310
-- config_name: triplets
-  features:
-  - name: qid
-    dtype: int64
-  - name: pos_pid
-    dtype: int64
-  - name: pos_score_original
-    dtype: float64
-  - name: pos_score_reranker
-    dtype: float64
-  - name: neg_count
-    dtype: int64
-  - name: neg_1_pid
-    dtype: int64
-  - name: neg_1_score_original
-    dtype: float64
-  - name: neg_1_score_reranker
-    dtype: float64
-  - name: neg_2_pid
-    dtype: int64
-  - name: neg_2_score_original
-    dtype: float64
-  - name: neg_2_score_reranker
-    dtype: float64
-  - name: neg_3_pid
-    dtype: int64
-  - name: neg_3_score_original
-    dtype: float64
-  - name: neg_3_score_reranker
-    dtype: float64
-  - name: neg_4_pid
-    dtype: int64
-  - name: neg_4_score_original
-    dtype: float64
-  - name: neg_4_score_reranker
-    dtype: float64
-  - name: neg_5_pid
-    dtype: int64
-  - name: neg_5_score_original
-    dtype: float64
-  - name: neg_5_score_reranker
-    dtype: float64
-  - name: neg_6_pid
-    dtype: int64
-  - name: neg_6_score_original
-    dtype: float64
-  - name: neg_6_score_reranker
-    dtype: float64
-  - name: neg_7_pid
-    dtype: int64
-  - name: neg_7_score_original
-    dtype: float64
-  - name: neg_7_score_reranker
-    dtype: float64
-  - name: neg_8_pid
-    dtype: int64
-  - name: neg_8_score_original
-    dtype: float64
-  - name: neg_8_score_reranker
-    dtype: float64
-  - name: neg_9_pid
-    dtype: int64
-  - name: neg_9_score_original
-    dtype: float64
-  - name: neg_9_score_reranker
-    dtype: float64
-  - name: neg_10_pid
-    dtype: int64
-  - name: neg_10_score_original
-    dtype: float64
-  - name: neg_10_score_reranker
-    dtype: float64
-  - name: neg_11_pid
-    dtype: int64
-  - name: neg_11_score_original
-    dtype: float64
-  - name: neg_11_score_reranker
-    dtype: float64
-  - name: neg_12_pid
-    dtype: int64
-  - name: neg_12_score_original
-    dtype: float64
-  - name: neg_12_score_reranker
-    dtype: float64
-  - name: neg_13_pid
-    dtype: int64
-  - name: neg_13_score_original
-    dtype: float64
-  - name: neg_13_score_reranker
-    dtype: float64
-  - name: neg_14_pid
-    dtype: int64
-  - name: neg_14_score_original
-    dtype: float64
-  - name: neg_14_score_reranker
-    dtype: float64
-  - name: neg_15_pid
-    dtype: int64
-  - name: neg_15_score_original
-    dtype: float64
-  - name: neg_15_score_reranker
-    dtype: float64
-  - name: neg_16_pid
-    dtype: int64
-  - name: neg_16_score_original
-    dtype: float64
-  - name: neg_16_score_reranker
-    dtype: float64
-  - name: neg_17_pid
-    dtype: int64
-  - name: neg_17_score_original
-    dtype: float64
-  - name: neg_17_score_reranker
-    dtype: float64
-  - name: neg_18_pid
-    dtype: int64
-  - name: neg_18_score_original
-    dtype: float64
-  - name: neg_18_score_reranker
-    dtype: float64
-  - name: neg_19_pid
-    dtype: int64
-  - name: neg_19_score_original
-    dtype: float64
-  - name: neg_19_score_reranker
-    dtype: float64
-  - name: neg_20_pid
-    dtype: int64
-  - name: neg_20_score_original
-    dtype: float64
-  - name: neg_20_score_reranker
-    dtype: float64
-  - name: neg_21_pid
-    dtype: int64
-  - name: neg_21_score_original
-    dtype: float64
-  - name: neg_21_score_reranker
-    dtype: float64
-  - name: neg_22_pid
-    dtype: int64
-  - name: neg_22_score_original
-    dtype: float64
-  - name: neg_22_score_reranker
-    dtype: float64
-  - name: neg_23_pid
-    dtype: int64
-  - name: neg_23_score_original
-    dtype: float64
-  - name: neg_23_score_reranker
-    dtype: float64
-  - name: neg_24_pid
-    dtype: int64
-  - name: neg_24_score_original
-    dtype: float64
-  - name: neg_24_score_reranker
-    dtype: float64
-  - name: neg_25_pid
-    dtype: int64
-  - name: neg_25_score_original
-    dtype: float64
-  - name: neg_25_score_reranker
-    dtype: float64
-  - name: neg_26_pid
-    dtype: int64
-  - name: neg_26_score_original
-    dtype: float64
-  - name: neg_26_score_reranker
-    dtype: float64
-  - name: neg_27_pid
-    dtype: int64
-  - name: neg_27_score_original
-    dtype: float64
-  - name: neg_27_score_reranker
-    dtype: float64
-  - name: neg_28_pid
-    dtype: int64
-  - name: neg_28_score_original
-    dtype: float64
-  - name: neg_28_score_reranker
-    dtype: float64
-  - name: neg_29_pid
-    dtype: int64
-  - name: neg_29_score_original
-    dtype: float64
-  - name: neg_29_score_reranker
-    dtype: float64
-  - name: neg_30_pid
-    dtype: int64
-  - name: neg_30_score_original
-    dtype: float64
-  - name: neg_30_score_reranker
-    dtype: float64
-  - name: neg_31_pid
-    dtype: int64
-  - name: neg_31_score_original
-    dtype: float64
-  - name: neg_31_score_reranker
-    dtype: float64
-  splits:
-  - name: train
-    num_bytes: 2510572624
-    num_examples: 3202261
-  download_size: 1205517229
-  dataset_size: 2510572624
+language:
+- az
+- en
+license: ms-pl
+task_categories:
+- text-retrieval
+tags:
+- retrieval
+- azerbaijani
+- information-retrieval
+- hard-negatives
+- reranker
+- ms-marco
+- dense-retrieval
+- colbert
+- bi-encoder
+- translated
+size_categories:
+- 1M<n<10M
 configs:
 - config_name: corpus
   data_files:
@@ -242,3 +32,153 @@ configs:
   - split: train
     path: triplets/train-*
 ---
+
+# MS MARCO Azerbaijani — Reranked Retrieval Training Dataset
+
+A large-scale passage retrieval training dataset in Azerbaijani, built by translating a 3.2M subset of the [MS MARCO](https://microsoft.github.io/msmarco/) passage ranking dataset and rescoring all query-passage pairs with a multilingual cross-encoder reranker.
+
+## Overview
+
+| | Count |
+|---|---|
+| Passages | 8,473,865 |
+| Queries | ~800,000 |
+| Triplets | 1,304,669 |
+| Negatives per triplet | up to 31 |
+| Total pairs scored | 41,746,530 |
+
+## Dataset Configs
+
+The dataset consists of three configs:
+
+### `corpus`
+
+The full translated passage collection.
+
+| Column | Type | Description |
+|---|---|---|
+| `pid` | int | Passage ID (original MS MARCO pid) |
+| `passage` | string | Passage text translated to Azerbaijani |
+
+### `queries`
+
+Translated queries.
+
+| Column | Type | Description |
+|---|---|---|
+| `qid` | int | Query ID (original MS MARCO qid) |
+| `query` | string | Query text translated to Azerbaijani |
+
+### `triplets`
+
+Training triplets with both original MS MARCO scores and reranker scores computed on the Azerbaijani translations.
+
+| Column | Type | Description |
+|---|---|---|
+| `qid` | int | Query ID (links to `queries`) |
+| `pos_pid` | int | Positive passage ID (links to `corpus`) |
+| `pos_score_original` | float | Original MS MARCO cross-encoder score (English) |
+| `pos_score_reranker` | float | Reranker score on Azerbaijani translation |
+| `neg_count` | int | Number of valid negatives for this triplet |
+| `neg_{k}_pid` | int | Passage ID of the k-th hard negative |
+| `neg_{k}_score_original` | float | Original MS MARCO score of the k-th negative |
+| `neg_{k}_score_reranker` | float | Reranker score of the k-th negative (Azerbaijani) |
+
+Negatives are sorted by `score_reranker` descending (hardest first). Columns run from `neg_1_*` to `neg_31_*`.
+
+## Construction Pipeline
+
+1. **Sampling**: 3.2M triplets were sampled from the MS MARCO `examples.json` using reservoir sampling, with 31 negatives selected per query
+2. **Translation**: All queries and passages were translated from English to Azerbaijani
+3. **Reranking**: Every query-passage pair (positive + all negatives) was scored with [BAAI/bge-reranker-v2-m3](https://huggingface.co/BAAI/bge-reranker-v2-m3) on the Azerbaijani translations (~14 hours, 41.7M pairs scored)
+4. **Output**: Triplets with dual scores (original English + Azerbaijani reranker) to enable flexible filtering during training
+
+## Why Reranker Scores?
+
+The original MS MARCO scores were computed on English text. After translation, semantic relationships between queries and passages can shift — some negatives become closer to the positive, and some positives become weaker. The reranker scores on Azerbaijani text reflect what the model will actually see during training.
+
+This also enables **false negative filtering**: negatives with `score_reranker > threshold * pos_score_reranker` are likely correct answers that MS MARCO did not annotate. These can be filtered out during training to avoid noisy supervision signals.
+
+## Usage
+
+```python
+from datasets import load_dataset
+
+corpus = load_dataset("LocalDoc/msmarco-az-reranked", "corpus")["train"]
+queries = load_dataset("LocalDoc/msmarco-az-reranked", "queries")["train"]
+triplets = load_dataset("LocalDoc/msmarco-az-reranked", "triplets")["train"]
+
+# Build lookups
+passage_lookup = {row["pid"]: row["passage"] for row in corpus}
+query_lookup = {row["qid"]: row["query"] for row in queries}
+
+# Inspect a triplet
+t = triplets[0]
+print(f"Query: {query_lookup[t['qid']]}")
+print(f"Positive [reranker={t['pos_score_reranker']:.4f}]: {passage_lookup[t['pos_pid']][:200]}")
+
+for k in range(1, 4):
+    neg_pid = t[f"neg_{k}_pid"]
+    neg_score = t[f"neg_{k}_score_reranker"]
+    if neg_pid:
+        print(f"Neg-{k} [reranker={neg_score:.4f}]: {passage_lookup[neg_pid][:200]}")
+```
+
+### Training with False Negative Filtering
+
+```python
+# Filter out false negatives where negative score > 95% of positive score
+FN_THRESHOLD = 0.95
+
+t = triplets[0]
+pos_score = t["pos_score_reranker"]
+cutoff = FN_THRESHOLD * pos_score
+
+clean_negs = []
+for k in range(1, 32):
+    neg_pid = t[f"neg_{k}_pid"]
+    neg_score = t[f"neg_{k}_score_reranker"]
+    if neg_pid and neg_score < cutoff:
+        clean_negs.append((neg_pid, neg_score))
+
+print(f"Original negatives: {t['neg_count']}")
+print(f"After FN filtering: {len(clean_negs)}")
+```
+
+## Example Output
+
+```
+Query: Dişi aslanlar nə qədər doğurur
+
+Positive [original=10.41, reranker=5.64]:
+Dişi şir normalda hər 18-26 aydan bir doğur. Təxminən 100-119 günlük
+hamiləlik dövründən sonra bir-altı bala doğur. Lakin, balaların sayı
+adətən üç və ya dörd olur və hər birinin çəkisi təxminən 3 funt olur.
+
+Neg-1 [original=9.26, reranker=7.41]:  ← false negative (reranker > positive)
+Dişi aslanlar adətən hər iki ildən bir bala doğurlar. Dişilər hamilə
+və ya əmizdirən deyillərsə, ildə bir neçə dəfə cütləşməyə hazırdırlar.
+
+Neg-2 [original=9.35, reranker=5.41]:
+Pride-ın dişi hissəsi bütün yetkinlik həyatlarını birlikdə yaşayır,
+lakin erkəklər gəlib-gedir. Dişi aslanın hamiləliyi təxminən dörd ay
+davam edir.
+
+Neg-3 [original=3.27, reranker=2.77]:  ← true negative
+At: Dişilərin hamiləliyi adətən 11-12 ay çəkir. Dəniz aslanı: Dəniz
+şirləri də balalarını 11-12 aylıq hamiləlik dövründən sonra dünyaya
+gətirirlər.
+```
+
+
+## Limitations
+
+- Passages and queries are machine-translated; translation artifacts (lexical mismatch, semantic drift) may affect quality
+- Reranker scores are from a multilingual model that may underperform on Azerbaijani compared to English
+- Original MS MARCO annotations are incomplete — some "negatives" are actually relevant (false negatives)
+
+
+
+## Contact
+
+For questions or issues, please contact LocalDoc at [v.resad.89@gmail.com].