ATD-MCL: Arukikata Travelogue Dataset with Geographic Entity Mention, Coreference, and Link Annotation

How to Restore the ATD-MCL Data

Install necessary Python libraries, for example, using pip install -r requirements.txt.
Obtain the Arukikata Travelogue Dataset (ATD) original data (data.zip) from the NII IDR site https://www.nii.ac.jp/dsc/idr/arukikata/.
Decompress data.zip and then move data directory to under atd directory (or create a symbolic link to data directory in atd directory).
Excute bin/gen_full_data_json.sh.
- The restored data will be placed at atd-mcl/full/main/json_per_doc/ and atd-mcl/full/main/split-*/json.
- The data used for calculating inter-annotator aggreement scores will be placed at atd-mcl/full/agreement/.
Excute bin/gen_full_data_tsv.sh.
- The restored data will be placed at atd-mcl/full/main/link_tsv_per_doc and atd-mcl/full/main/mention_tsv_per_doc.

Data Statistics

The Set-A data consists of 100 documents annotated only with mention and coreference information. The Set-B data consists of 100 documents annotated with mention, coreference, and link information. (An entity corresponds to a coreference cluster of mentions.)

	#Doc	#Sent	#Word	#Mention	#Entity
Set-A	100	5,949	85,741	6,052	3,131
Set-B	100	6,324	87,074	6,119	3,208
Total	200	12,273	172,815	12,171	6,339

Data Split

See docs/data_split.md.

Data Format

JSON Data Format

The JSON data (atd-mcl/full/main/split-*/json and atd-mcl/full/main/json_per_doc) holds full annotation information as follows.

A document object value is assosiated with a key that represents the document ID (e.g., 00019). Each document object has the sets of sections, sentences, mentions, and entities.

 {
   "00019": {
     "sections": {
       "001": {
       ...
       },
     },
     "sentences": {
       "001": {
       ...
       },
     },
     "mentions": {
       "M001": {
         ...
       },
     },
     "entities": {
       "E001": {
         ...
       }
     }
   }
 }

A section object under sections is as follows:

"sections": {
  "001": {
    "sentence_ids": [
      "001",
      "002",
      "003"
    ]
  },
...

A sentence object under sentences is as follows:

A sentence object may have one or more geographic entity mentions.
Some sentences with an ID that has a branch number (e.g., "026-01" and "026-02") indicate that a line of text in the original ATD data was split into those multiple sentences.

"sentences": {
  "001": {
    "section_id": "001",
    "text": "奈良公園のアイドル「しか」で~す。",
    "mention_ids": [
      "M001"
    ]
  },
  ...
  "026-01": {
    "section_id": "013",
    "span_in_orig_text": [
      0,
      8
    ],
    "text": "とにかく広~い!",
    "mention_ids": []
  },
  "026-02": {
    "section_id": "013",
    "span_in_orig_text": [
      8,
      16
    ],
    "text": "そして静かです。",
    "mention_ids": []
  },

A mention object under mentions is as follows:

A mention object may be associated with an entity.

"mentions": {
  "M001": {
    "sentence_id": "001",
    "entity_id": "E001",
    "span": [
      0,
      4
    ],
    "text": "奈良公園",
    "entity_type": "FAC_NAME"
  },

An entity object, which corresponds to a coreference cluster of one or more mentions, under entities is as follows:

An entity object is associated with one or more mentions.
has_name indicates whether at least one member mention's entity type is *_NAME or not.

"entities": {
  "E001": {
    "original_entity_id": "C1",
    "entity_type_merged": "FAC",
    "has_name": true,
    "has_reference": true,
    "best_ref_type": "OSM",
    "best_ref_url": "https://www.openstreetmap.org/way/456314269",
    "best_ref_query": "奈良公園",
    "member_mention_ids": [
      "M001",
      "M011"
    ]
  },

Mention TSV Data Format

The mention TSV data (atd-mcl/full/main/mention_tsv_per_doc) holds mention-related annotation information as follows.

1st column: document_id
2nd column: section_id:sentence_id
3rd column: Sentence text
4th column: Mention information with the following elements. Multiple mentions are enumerated with ";".
- 1st element: mention_id
- 2nd element: span
- 3rd element: entity_type
- 4th element: mention text
- 5th element: entity_id
- 6th element: generic
- 7th element: ref_spec_amb
- 8th element: ref_hie_amb

Example:

00019	002:004	奈良の有名スポットですよね!	M002,0:2,LOC_NAME,奈良,E002,,,ref_hie_amb;M011,3:9,LOC_OR_FAC,有名スポット,E001,,,

Link TSV Data Format

The link TSV data (atd-mcl/full/main/link_tsv_per_doc) holds link-related annotation information. Specifically, entities and their member mentions (except for GENERIC and SPEC_AMB entities/mentions) are listed in TSV rows. The column with a non-empty entity_id value corresponds to an entity, and the column with a non-empty mention_id value corresponds to a member mention of the preceding entity column.

Example:

#document_id	entity_id	mention_id	best_ref_type	best_ref_url	best_ref_query	best_ref_is_overseas	second_A_ref_type	second_A_ref_url	second_A_ref_query	second_A_ref_is_overseas	second_B_ref_type	second_B_ref_url	second_B_ref_query	second_B_ref_is_overseas	entity_type	span	normalized_name	mention_text	ref_hie_amb	sentence_id	sentence_text
00019	E002	-	OSM	https://www.openstreetmap.org/relation/3227707	奈良市		OSM	https://www.openstreetmap.org/relation/358631	奈良県						LOC	-	奈良	-	-
00019	-	E002:M002	-	-	-	-	-	-	-	-	-	-	-	-	LOC_NAME	0:2	-	奈良	ref_hie_amb	002:004	奈良の有名スポットですよね!

Notes:

mention_id column values acutally represent "entity_id:mention_id".
sentence_id column values acutally represent "section_id:sentence_id".

Detailed Data Specification

See docs/data_specification.

Contact

Shohei Higashiyama <shohei.higashiyama [at] nict.go.jp>
Hiroki Ouchi <hiroki.ouchi [at] is.naist.jp>

Acknowledgements

This study was supported by JSPS KAKENHI Grant Number JP22H03648. The annotation data was constructed by IR-Advanced Linguistic Technologies Inc.

Citation

Please cite the following paper.

@inproceedings{higashiyama-etal-2024-arukikata,
    title = "Arukikata Travelogue Dataset with Geographic Entity Mention, Coreference, and Link Annotation",
    author = "Higashiyama, Shohei and Ouchi, Hiroki and Teranishi, Hiroki and Otomo, Hiroyuki and Ide, Yusuke and Yamamoto, Aitaro and Shindo, Hiroyuki and Matsuda, Yuki and Wakamiya, Shoko and Inoue, Naoya and Yamada, Ikuya and Watanabe, Taro",
    booktitle = "Findings of the Association for Computational Linguistics: EACL 2024",
    month = mar,
    year = "2024",
}

naist-nlp / atd-mcl