Add corpus, queries and pseudo-label pairs (parquet) + dataset card

README.md

@@ -0,0 +1,230 @@
+---
+license: cc-by-sa-4.0
+language:
+- zh
+- en
+task_categories:
+- text-retrieval
+tags:
+- multimodal-retrieval
+- traditional-chinese
+- cross-lingual
+- wikipedia
+- image-captioning
+pretty_name: Mr. Right zh-TW (Traditional Chinese Multimodal Retrieval Corpus)
+size_categories:
+- 100K<n<1M
+configs:
+- config_name: corpus
+  data_files:
+  - split: train
+    path: corpus/train-*.parquet
+- config_name: queries-test
+  data_files:
+  - split: test
+    path: queries_test/train-00000.parquet
+- config_name: queries-val
+  data_files:
+  - split: validation
+    path: queries_val/train-00000.parquet
+- config_name: queries-mt-en
+  data_files:
+  - split: train
+    path: queries_mt_en/train-00000.parquet
+- config_name: pseudo-pairs-zh
+  data_files:
+  - split: train
+    path: pseudo_pairs_zh/train-00000.parquet
+- config_name: pseudo-pairs-hard
+  data_files:
+  - split: train
+    path: pseudo_pairs_hard/train-00000.parquet
+---
+
+# Mr. Right zh-TW — Traditional-Chinese Multimodal Retrieval Corpus
+
+A **bilingual (Traditional Chinese + English) multimodal document retrieval** corpus:
+769,245 Wikipedia-derived documents, each with an English title/body, a Traditional-Chinese
+(Taiwan) translation, and a **machine-generated Traditional-Chinese caption of the document's
+image** — plus test/validation query sets and the pseudo-labelled pairs used to fine-tune a
+retriever on it.
+
+This is a derivative of the [Mr. Right](https://github.com/hsiehjackson/Mr.Right) collection
+(Hsieh et al., 2022), translated into Traditional Chinese and enriched with image captions.
+To our knowledge it is the first Traditional-Chinese multimodal IR corpus of this scale on the Hub.
+
+**Images are not redistributed here.** Every row carries `img_url` (Wikimedia) and `sha1`, so
+the image tree is reproducible; see [Getting the images](#getting-the-images).
+
+Companion repositories:
+- Retriever adapters: [`ericssonbear/qwen3-vl-emb-8b-mrright-zhtw-lora`](https://huggingface.co/ericssonbear/qwen3-vl-emb-8b-mrright-zhtw-lora)
+- Pre-computed document embeddings: [`ericssonbear/mr-right-zhtw-embeddings`](https://huggingface.co/datasets/ericssonbear/mr-right-zhtw-embeddings)
+
+---
+
+## ⚠️ Read this before you evaluate: the `query_multi` image trap
+
+The upstream `multi` query type ships **with a query image**. We found that **1,992 of ~2,000**
+test queries use the *same image file* as their gold document. A completely untrained backbone
+scores **Recall@10 = .952** on that field — it is measuring image hashing, not retrieval.
+
+**Use `query_multi` / `query_multi_zhtw` as plain text and do not feed the query image.**
+All numbers in the companion model card are produced under this text-only protocol.
+If you skip this, your results will be systematically inflated and incomparable.
+
+Second protocol note — **the clean subset**: 96 of the 2,047 test queries point at documents
+whose image URL is a dead link, so those documents are absent from the corpus and are
+unreachable by any system. Scoring on all 2,047 imposes a recall ceiling of .953. We report on
+the **clean** subset (gold document present in the corpus, n = 1,951).
+
+---
+
+## Configs
+
+| config | rows | what it is |
+|---|---:|---|
+| `corpus` | 769,245 | the documents (English + zh-TW text, zh-TW image caption, image URL) |
+| `queries-test` | 2,047 | test queries, 3 types × 2 languages |
+| `queries-val` | 100 | validation queries, same schema |
+| `queries-mt-en` | 2,147 | machine-translated English `query_multi` for test+val (cross-lingual baseline) |
+| `pseudo-pairs-zh` | 32,359 | pseudo-labelled training pairs (round-trip ranked) |
+| `pseudo-pairs-hard` | 27,055 | hard-negative-mined pairs with an LLM judge score |
+
+```python
+from datasets import load_dataset
+
+corpus  = load_dataset("ericssonbear/mr-right-zhtw", "corpus",       split="train")
+queries = load_dataset("ericssonbear/mr-right-zhtw", "queries-test", split="test")
+```
+
+### `corpus` schema
+
+| field | type | notes |
+|---|---|---|
+| `id` | int64 | upstream Mr. Right document id — **non-contiguous** (37,112 of 806,357 dropped) |
+| `title`, `doc_text` | string | English original |
+| `title_zhtw`, `doc_text_zhtw` | string | Traditional-Chinese translation (upstream `g4` split) |
+| `img_caption` | string | zh-TW caption of the image, machine-generated |
+| `img_url` | string | Wikimedia URL of the image |
+| `caption_source` | string | `own` (187,057) or `external` (582,188) — see below |
+| `sha1` | string | SHA-1 of the image bytes; stable join key since `id` is gappy |
+| `image_path` | string | `<sha1[:2]>/<sha1>.<ext>` — the content-addressed layout to reproduce locally |
+
+### `queries-*` schema
+
+`id` (the gold document), `img_url`, `title`, `doc_text`, and three query types in both
+languages: `query_img` / `query_text` / `query_multi` and their `_zhtw` counterparts.
+`id` joins to `corpus.id`; relevance is one gold document per query.
+
+### `pseudo-pairs-*` schema
+
+`id` (gold document), `query_multi_zhtw`, `query_multi`, `rt_rank_zh`, `rt_rank_en`
+(round-trip rank of the gold document under the zh/en query — lower is a cleaner pair;
+`null` = not retrieved), and for `hard`, `judge_score` (LLM judge, higher is better).
+
+---
+
+## How it was built
+
+```
+Wikipedia (CC BY-SA)
+   └─ Mr. Right collection (Hsieh et al. 2022, CC BY-SA 4.0) — 806,357 docs, English
+        └─ g4 split: Traditional-Chinese translation of title + doc_text   [upstream]
+             └─ images fetched from Wikimedia by img_url  →  SHA-1 content-addressed tree
+                  └─ zh-TW caption per image
+                  └─ keep a doc iff its image resolved AND its caption passed the quality gate
+                       →  769,245 docs  (this dataset)
+```
+
+**Captions.** Two sources, recorded per row in `caption_source`:
+- `own` (187,057 docs) — generated for this project with **Qwen3-VL-8B-Instruct** under vLLM,
+  greedy decoding, zh-TW instruction prompt, 256 max tokens.
+- `external` (582,188 docs) — three caption sets produced by other members of the same lab
+  with their own VLM pipelines, joined by document id.
+
+**Quality gate.** A document is kept only if its image was successfully fetched *and* its
+caption is non-empty, contains no replacement characters, and has at least 10 CJK characters.
+37,112 documents were dropped (dead image links or failed captions) — hence the gappy `id`.
+
+## Getting the images
+
+The image tree is ~1.2 TB / 769,217 files, so it is not redistributed. Rebuild it from
+`img_url` and verify each file against `sha1`:
+
+```python
+import hashlib, pathlib, requests
+from datasets import load_dataset
+
+UA = "your-project/1.0 (your-contact@example.com)"   # Wikimedia requires a real UA
+root = pathlib.Path("images")
+
+for row in load_dataset("ericssonbear/mr-right-zhtw", "corpus", split="train", streaming=True):
+    dst = root / row["image_path"]
+    if dst.exists():
+        continue
+    r = requests.get(row["img_url"], headers={"User-Agent": UA}, timeout=30)
+    r.raise_for_status()
+    assert hashlib.sha1(r.content).hexdigest() == row["sha1"], row["id"]
+    dst.parent.mkdir(parents=True, exist_ok=True)
+    dst.write_bytes(r.content)
+```
+
+Wikimedia rate-limits to roughly **1 image/s per egress IP** and expects a descriptive
+User-Agent and honoured `Retry-After`. A full rebuild from one IP takes on the order of
+9 days; plan accordingly. Content addressing means parallel downloaders cannot collide.
+
+If you only need the document vectors, skip the images and use the
+[pre-computed embeddings](https://huggingface.co/datasets/ericssonbear/mr-right-zhtw-embeddings).
+
+## Limitations
+
+- **Captions are machine-generated** and inherit VLM failure modes (hallucinated detail, OCR
+  errors, occasional English leakage). They were filtered for form, not for factual accuracy.
+- **`doc_text_zhtw` is machine-translated**, not human-authored or human-reviewed. Spot checks
+  found residual Chinese/English mixing in a small fraction of translated queries (~2.6% in the
+  MT English query set).
+- **No human-authored Chinese evaluation slice.** The zh-TW queries are translations, so they
+  carry translationese; results on them are not a substitute for native-speaker queries.
+- `caption_source` is confounded with document id range (`own` covers the high-id tail), so
+  it should not be read as a controlled comparison between captioning pipelines.
+- The corpus is a public encyclopedia: no personal data beyond what Wikipedia already
+  publishes, and no human-subject research is involved.
+
+## License and attribution
+
+Released under **CC BY-SA 4.0**, inherited from the upstream Mr. Right collection, which in
+turn derives from Wikipedia (CC BY-SA). Any redistribution or adaptation must stay under
+CC BY-SA 4.0 and credit both this dataset and the upstream authors.
+
+Individual images remain under their own Wikimedia licenses and are **not** covered by this
+dataset's license — that is one reason only URLs are shipped here.
+
+Upstream:
+
+```bibtex
+@article{hsieh2022mrright,
+  title  = {Mr. Right: Multimodal Retrieval on Representation of ImaGe witH Text},
+  author = {Hsieh, Cheng-Ping and Chen, Jui-Ting and others},
+  journal = {arXiv preprint arXiv:2209.13764},
+  year   = {2022}
+}
+```
+
+## Citation
+
+<!-- TODO(authors): replace the placeholder author list below before making this repo public. -->
+
+```bibtex
+@misc{mrright_zhtw_2026,
+  title  = {Mr. Right zh-TW: A Traditional-Chinese Multimodal Retrieval Corpus},
+  author = {AUTHORS_PLACEHOLDER},
+  year   = {2026},
+  url    = {https://huggingface.co/datasets/ericssonbear/mr-right-zhtw}
+}
+```
+
+## Acknowledgements
+
+Built by AUTHORS_PLACEHOLDER. Heavy compute ran on the NCHC 晶創 (Nano5) cluster via Slurm.
+Image captioning for the low-id range was contributed by lab members running their own
+VLM pipelines. Funded under FUNDING_PLACEHOLDER.

corpus/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a29bbe74fd11ff2003076ede09592e5ed96670362ca5f5abaad7a17aae35426b
+size 83870001

corpus/train-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e8f5a930d0f22130a798a8aa40f015793554c824cb326c4f783143748d8dcf8
+size 83650339

corpus/train-00002.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:446bb96d681b71557b06fcbf0caa90f9dbf87704aa5f4d657f9020c6a43ff2ba
+size 83812705

corpus/train-00003.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d4530a1f7d01619115d44acd5ba60109aa94230c05a61bf5725338343e555034
+size 83741894

corpus/train-00004.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da250bdf89f7c5a33e315c6d47caceff17b98d7500bf1bc5a60747124d8d6221
+size 81531989

corpus/train-00005.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b171410736e507abef0845b9b877387b0bfea1ff7624d5df07f0020f6fd61820
+size 75750273

corpus/train-00006.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:738697b5ad9b3c77b1c5d1cb4da972b1350daa9a34bca27987908b2b8ee78261
+size 52894231

corpus/train-00007.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61dd91c99a519375f717ab6c2ce86d3f4a82a4dbaff0ad617338da15f60ab26c
+size 36709221

pseudo_pairs_hard/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0a4263bd21a9da9af43d1ec90b512ce9991c099b104cea9ac32c3d4ff309a35
+size 1705260

pseudo_pairs_zh/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ea4a4d019934f25231fc0a64a7e4966b014abc3c76db2f2792997dddca6f81a
+size 2856850

queries_mt_en/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45ffa786bcd28cf39243a2171834483d96824ddeb0320e8df8f698918b4b1f68
+size 90477

queries_test/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6c46dbb4f643412bcf83bb67fb1888d42828682a0284c258cc8c7eb3a5ab0e2
+size 1209438

queries_val/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6d52e5e8e61f327bb6e8cf758c38ab7dbba2773cf20fbb075f68e1bd67f1c4da
+size 30837