About the Challenge

Tabular data in the form of CSV files is the common input format in a data analytics pipeline. However, a lack of understanding of the semantic structure and meaning of the content may hinder the data analytics process. Thus gaining this semantic understanding will be very valuable for data integration, data cleaning, data mining, machine learning and knowledge discovery tasks. For example, understanding what the data is can help assess what sorts of transformation are appropriate on the data.

Tables on the Web may also be the source of highly valuable data. The addition of semantic information to Web tables may enhance a wide range of applications, such as web search, question answering, and knowledge base (KB) construction.

Tabular data to Knowledge Graph (KG) matching is the process of assigning semantic tags from Knowledge Graphs (e.g., Wikidata or DBpedia) to the elements of the table. This task however is often difficult in practice due to metadata (e.g., table and column names) being missing, incomplete or ambiguous.

The SemTab challenge aims at benchmarking systems dealing with the tabular data to KG matching problem, so as to facilitate their comparison on the same basis and the reproducibility of the results.

The 2022 edition of this challenge will be collocated with the 21st International Semantic Web Conference and the 17th International Workshop on Ontology Matching.

Participation: Forum and Registration

We have a discussion group for the challenge where we share the latest news with the participants and we discuss issues risen during the evaluation rounds.

Please register your system using this google form.

Note that participants can join SemTab at any Round for any of the tasks/tracks.

Challenge Tracks

Accuracy Track

The evaluation of systems regarding accuracy is similar to prior versions of the SemTab.
That is, to illustrate the accuracy of the submissions, we evaluate systems on typical multi-class classification metrics as detailed below.
In addition, we adopt the "cscore" for the CTA task to reflect the distance in the type hierarchy between the predicted column type and the ground truth semantic type.


Matching Tasks:
  • CTA Task: Assigning a semantic type (a DBpedia class as fine-grained as possible) to a column.
  • CEA Task: Matching a cell to a Wikidata entity.
  • CPA Task: Assigning a KG property to the relationship between two columns.
Matching Criteria:
  • Average Precision
  • Average Recall
  • Average F1
  • Cscore
Important Dates (all 2022):
  • May 26: First call for challenge participants.
  • June 13 - July 13: Round 1.
  • July 15 - August 14: Round 2.
  • August 15: Inivations to present at the ISWC conference.
  • September 15 - October 15: Round 3.
  • October 21: Paper submissions (via easychair), and artifact publication.
  • October 23 - 27: Challenge presentation during OM workshop.
  • October 23 - 27: Challenge Presentation and prize announcement during ISWC.
  • November 15: Final version papers (via easychair).

Datasets Track

The data that table-to-Knowledge-Graph matching systems are trained and evaluated on, is critical for their accuracy and relevance.
We invite dataset submissions that provide challenging and accessible new datasets to advance the state-of-the-art of table-to-KG matching systems.
Preferably, these datasets provide tables along with their ground truth annotations for at least one of CEA, CTA and CPA tasks.
The dataset may be general or specific to a certain domain.

Submissions will be evaluated according to provide the following:
  • Description of the data collection, curation, and annotation processes.
  • Availability of documentation with insights in the dataset content.
  • Publicly accessible link to the dataset (e.g. Zenodo) and its DOI.
  • Explanation of maintenance and long-term availability.
  • Clear description of the envisioned use-cases.
  • Application in which the dataset is used to solve an exemplar task.
We ask participants to describe their datasets submissions via easychair in a short paper (max 6 pages) that discusses how the above criteria are covered, while also including a link to the resources. The link to the resources may be private, until the submission is evaluated by the SemTab organisers. See paper guidelines below, for more details. More guidance for creating, documenting and publishing datasets can be found here.

Important Dates (all 2022, tentative):
  • August 15: Paper submissions (via easychair), and artifact publication.
  • September 30: Notification of accept/reject.
  • October 23 - 27: Dataset Presentation and prize announcement during ISWC.
  • November 15: Final version papers (via easychair).

Artifacts Availability Badge

New this year is the Artifacts Availability Badge which is applicable to the Accuracy Track as well as the Datasets Track.
The goal of this badge is to motivate authors to publish and document their systems, code, and data, so that others can use these artifacts and potentially reproduce or build on the results.
This badge is given if all resources are verified to satisfy the below criteria.

The criteria used to assess submissions (both accuracy and dataset submissions) are:
  • Publicly accessible data (if applicable).
  • Publicly accessible source code.
  • Clear documentation of the code and data.
  • Open-source dependencies.

Datasets and tasks per round

Round #1

Column Type Annotation by Wikidata (CTA-WD)

This is a task of ISWC 2022 "Semantic Web Challenge on Tabular Data to Knowledge Graph Matching". It's to annotate an entity column (i.e., a column composed of entity mentions) in a table with types from Wikidata (version: 20220521)
Notes: participants may use the public Wikidata endpoint (or its API) since the above dump is very recent.

Task Description

The task is to annotate each entity column by items of Wikidata as its type. Each column can be annotated by multiple types: the one that is as fine grained as possible and correct to all the column cells, is regarded as a perfect annotation; the one that is the ancestor of the perfect annotation is regarded as an okay annotation; others are regarded as wrong annotations.

The annotation can be a normal entity of Wikidata, with the prefix of http://www.wikidata.org/entity/, such as http://www.wikidata.org/entity/Q8425. Each column should be annotated by at most one item. A perfect annotation is encouraged with a full score, while an okay annotation can still get a part of the score. Example: "KIN0LD6C","0","http://www.wikidata.org/entity/Q8425". Please use the prefix of http://www.wikidata.org/entity/ instead of the URL prefix https://www.wikidata.org/wiki/.

The annotation should be represented by its full IRI, where the case is NOT sensitive. Each submission should be a CSV file. Each line should include a column identified by table id and column id, and the column's annotation (a Wikidata item). It means one line should include three fields: "Table ID", "Column ID" and "Annotation IRI". The headers should be excluded from the submission file.

Notes:
  1. Table ID is the filename of the table data, but does NOT include the extension.
  2. Column ID is the position of the column in the input, starting from 0, i.e., first column's ID is 0.
  3. One submission file should have NO duplicate lines for each target column.
  4. Annotations for columns out of the target columns are ignored.
Dataset
Link
Round #1 Dataset
Description

The dataset contains:

  • evaluator codes (CTA_WD_Evaluator.py)
  • the validation set (DataSets/HardTablesR1/Valid/gt/cta_gt.csv, DataSets/HardTablesR1/Valid/gt/cta_gt_ancestor.json, DataSets/HardTablesR1/Valid/gt/cta_gt_descendent.json, DataSets/HardTablesR1/Valid/tables)
  • the testing set (DataSets/HardTablesR1/Test/tables, DataSets/HardTablesR1/Test/target/cta_gt.csv)
Format
One table is stored in one CSV file. Each line corresponds to a table row. The first row may either be the table header or content. The target columns for annotation are saved in a CSV file. The CTA GTs' ancestors and descendents are saved in two json files, respectively.
Evaluation Criteria

We encourage one perfect annotation, and at same time score one of its ancestors (okay annotation). Thus we calculate Approximate Precision (\(APrecision\)), Approximate Recall (\(ARecall\)), and Approximate F1 Score (\(AF1\)): \[APrecision = {\sum_{a \in all\ annotations}g(a) \over all\ annotations\ \#}\] \[ARecall = {\sum_{col \in all\ target\ columns}(max\_annotation\_score(col)) \over all\ target\ columns\ \#}\] \[AF1 = {2 \times APrecision \times ARecall \over APrecision + ARecall}\]

Notes:
  1. # denotes the number.
  2. \( g(a) = \begin{cases} 1.0, & \text{ if } a \text{ is a perfect annotation} \\ 0.8^{d(a)}, & \text{ if } a \text{ is an ancestor of the perfect annotation and } d(a) < 5 \\ 0.7^{d(a)}, & \text{ if } a \text{ is a descendent of the perfect annotation and } d(a) < 3 \\ 0, & otherwise \end{cases} \)

    where \(d(a)\) is the depth to the perfect annotation. E.g., \(d(a)=1\) if \(a\) is a parent of the perfect annotation, and \(d(a)=2\) if \(a\) is a grandparent of the perfect annotation.

  3. \( max\_annotation\_score(col) = \begin{cases} g(a), & \text{ if } col \text{ has an annotation } a \\ 0, & \text{ if } col \text{ has no annotation } \end{cases} \)
  4. \(AF1\) is used as the primary score, and \(APrecision\) is used as the secondary score.
  5. A cell may have multiple equivalent Wikidata items as its GT (e.g., redirected pages Q20514736 and Q852446). For an annotated entity, our evaluator will calculate the score with each GT entity and select the maximum score.
Submission
Participants can test and develop their systems on the given ground truth (validation set). They can weekly upload their annotations corresponding to the targets (test set).

Cell Entity Annotation by Wikidata (CEA-WD)

This is a task of ISWC 2022 "Semantic Web Challenge on Tabular Data to Knowledge Graph Matching". It is to annotate column cells (entity mentions) in a table with entities of Wikidata (version: 20220521)
Notes: participants may use the public Wikidata endpoint (or its API) since the above dump is very recent.

Task Description

The task is to annotate each target cell with an entity of Wikidata. Each submission should contain the annotation of the target cell. One cell can be annotated by one entity with the prefix of http://www.wikidata.org/entity/. Any of the equivalent entities of the ground truth entity are regarded as correct. Case is NOT sensitive.

The submission file should be in CSV format. Each line should contain the annotation of one cell which is identified by a table id, a column id and a row id. Namely one line should have four fields: "Table ID", "Row ID", "Column ID" and "Entity IRI". Each cell should be annotated by at most one entity. The headers should be excluded from the submission file. Here is an example: "OHGI1JNY","32","1","http://www.wikidata.org/entity/Q5484". Please use the prefix of http://www.wikidata.org/entity/ instead of https://www.wikidata.org/wiki/ which is the prefix of the Wikidata page URL.

Notes:
  1. Table ID does not include filename extension; make sure you remove the .csv extension from the filename.
  2. Column ID is the position of the column in the table file, starting from 0, i.e., first column's ID is 0.
  3. Row ID is the position of the row in the table file, starting from 0, i.e., first row's ID is 0.
  4. One submission file should have NO duplicate lines for one cell.
  5. Annotations for cells out of the target cells are ignored.
Dataset
Link
Round #1 Dataset
Description

The dataset contains:

  • evaluator codes (CEA_WD_Evaluator.py)
  • the validation set (DataSets/HardTablesR1/Valid/gt/cea_gt.csv, DataSets/HardTablesR1/Valid/tables)
  • the testing set (DataSets/HardTablesR1/Test/tables, DataSets/HardTablesR1/Test/target/cea_target.csv)
Format
One table is stored in one CSV file. Each line corresponds to a table row. The first row may either be the table header or content. The target cells for annotation are saved in a CSV file.
Evaluation Criteria

Precision, Recall and F1 Score are calculated: \[Precision = {{correct\_annotations \#} \over {submitted\_annotations \#}}\] \[Recall = {{correct\_annotations \#} \over {ground\_truth\_annotations \#}}\] \[F1 = {2 \times Precision \times Recall \over Precision + Recall}\]

Notes:
  1. # denotes the number.
  2. \(F1\) is used as the primary score, and \(Precision\) is used as the secondary score.
  3. One target cell, one ground truth annotation, i.e., # ground truth annotations = # target cells. The ground truth annotation has already covered all equivalent entities (e.g., wiki page redirected entities); the ground truth is hit if one of its equivalent entities is hit.
Submission
Participants can test and develop their systems on the given ground truth (validation set). They can weekly upload their annotations corresponding to the targets (test set).

Column Property Annotation by Wikidata (CPA-WD)

This is a task of ISWC 2022 "Semantic Web Challenge on Tabular Data to Knowledge Graph Matching". It is to annotate column relationships in a table with properties of Wikidata (version: 20220521)
Notes: participants may use the public Wikidata endpoint (or its API) since the above dump is very recent.

Task Description

The task is to annotate each column pair with a property of Wikidata. Each submission should contain an annotation of a target column pair. Note the order of the two columns matters. The annotation property should start with the prefix of http://www.wikidata.org/prop/direct/. Case is NOT sensitive.

The submission file should be in CSV format. Each line should contain the annotation of two columns which is identified by a table id, column id one and column id two. Namely one line should have four fields: "Table ID", "Column ID 1", "Column ID 2" and "Property IRI". Each column pair should be annotated by at most one property. The headers should be excluded from the submission file. Here is an example: "OHGI1JNY","0","1","http://www.wikidata.org/prop/direct/P702". Please use the prefix of http://www.wikidata.org/prop/direct/ instead of https://www.wikidata.org/wiki/ which is the prefix of the Wikidata page URL.

Notes:
  1. Table ID does not include filename extension; make sure you remove the .csv extension from the filename.
  2. Column ID is the position of the column in the table file, starting from 0, i.e., first column's ID is 0.
  3. One submission file should have NO duplicate lines for one column pair.
  4. Annotations for column pairs out of the targets are ignored.
Dataset
Link
Round #1 Dataset
Description

The dataset contains:

  • evaluator codes (CPA_WD_Evaluator.py))
  • the validation set (DataSets/HardTablesR1/Valid/gt/cpa_gt.csv, DataSets/HardTablesR1/Valid/tables)
  • the testing set (DataSets/HardTablesR1/Test/tables, DataSets/HardTablesR1/Test/target/cpa_target.csv)
Format
One table is stored in one CSV file. Each line corresponds to a table row. The first row may either be the table header or content. The target cells for annotation are saved in a CSV file.
Evaluation Criteria

Precision, Recall and F1 Score are calculated: \[Precision = {{correct\_annotations \#} \over {submitted\_annotations \#}}\] \[Recall = {{correct\_annotations \#} \over {ground\_truth\_annotations \#}}\] \[F1 = {2 \times Precision \times Recall \over Precision + Recall}\]

Notes:
  1. # denotes the number.
  2. \(F1\) is used as the primary score, and \(Precision\) is used as the secondary score.
  3. One target column pair, one ground truth annotation, i.e., # ground truth annotations = # target column pairs.
Submission
Participants can test and develop their systems on the given ground truth (validation set). They can weekly upload their annotations corresponding to the targets (test set).

Round #2

Round #3

Paper Guidelines

We invite participants in the Accuracy Track as well as the Datasets Track to submit a paper using easychair.
System papers in the Accuracy Track should be no more than 12 pages long (excluding references) and papers for the Datasets Track are limited to 6 pages.
If you are submitting to the Datasets Track, please append "[Datasets Track]" at the end of the paper title.
Both type of papers should be formatted using the CEUR Latex template or the CEUR Word template. Papers will be reviewed by 1-2 challenge organisers.

Accepted papers will be published as a volume of CEUR-WS. By submitting a paper, the authors accept the CEUR-WS publishing rules.

Organisation

This challenge is organised by Kavitha Srinivas (IBM Research), Ernesto Jiménez-Ruiz (City, University of London; University of Oslo), Oktie Hassanzadeh (IBM Research), Jiaoyan Chen (University of Oxford), Vasilis Efthymiou (FORTH - ICS), Vincenzo Cutrona (SUPSI), Juan Sequeda (data.world), Nora Abdelmageed (University of Jena), and Madelon Hulsebos (Sigma Computing, University of Amsterdam). If you have any problems working with the datasets or any suggestions related to this challenge, do not hesitate to contact us via the discussion group.

Acknowledgements

The challenge is currently supported by the SIRIUS Centre for Research-driven Innovation and IBM Research.