phucty / mtab_tool

MTab: Entity Search and Table Annotation with Wikidata, Wikipedia, and DBpedia

---

MTab: Entity Search and Table Annotation with Knowledge Graphs (Wikidata, Wikipedia and DBpedia)

Demo

• Demo (Teaser): https://youtu.be/sr-zxH5JUjw
• Demonstration Video: https://youtu.be/0ibTWeObWaA
• Entity Search: https://mtab.kgraph.jp/mtabes
• Table Annotation: https://mtab.kgraph.jp/mtab

API usage

• Entity Search: MTabES
  • Request Example: entity search
• Table Annotation: MTab
  • Request Example: table annotation
• Running Table Annotation on Tough Tables (2T) dataset:
  • Running Scripts: run 2t

Source code:

• Development code: https://github.com/phucty/mtab_dev

Other works:

• MTab4D: Table Annotation with DBpedia
• WikiDB: Build a DB (key-value store - LMDB style) from Wikidata dump, offline access Wikidata, fast boolean search

References

• Phuc Nguyen, Ikuya Yamada, Natthawut Kertkeidkachorn, Ryutaro Ichise, Hideaki Takeda, SemTab 2021: Tabular Data Annotation with MTab Tool. [video]

• Phuc Nguyen, Ikuya Yamada, Natthawut Kertkeidkachorn, Ryutaro Ichise, Hideaki Takeda, Demonstration of MTab: Tabular Data Annotation with Knowledge Graphs. [video]

• Phuc Nguyen, Hideaki Takeda, MTab: Tabular Data Annotation, NII Open House June 2021. [video]

• Phuc Nguyen, Ikuya Yamada, Hideaki Takeda, MTabES: Entity Search with Keyword Search, Fuzzy Search, and Entity Popularities, In The 35th Annual Conference of the Japanese Society for Artificial Intelligence (JSAI), 2021. [video]

• Phuc Nguyen, Ikuya Yamada, Natthawut Kertkeidkachorn, Ryutaro Ichise, Hideaki Takeda, MTab4Wikidata at SemTab 2020: Tabular Data Annotation with Wikidata, In SemTab@ISWC, 2020. [video]

• Phuc Nguyen, Natthawut Kertkeidkachorn, Ryutaro Ichise, Hideaki Takeda MTab: Matching Tabular Data to Knowledge Graph using Probability Models, In SemTab@ISWC, 2019, [slides]

Awards:

• 1st prize at SemTab 2021 (usability track). Results

  

• 1st prize at SemTab 2020 (tabular data to Wikidata matching). Results

• 1st prize at SemTab 2019 (tabular data to DBpedia matching). Results

Citing

If you find MTab tool useful in your work, and you want to cite our work, please use the following referencee:

@inproceedings{2021_mtab4wikidata,
  author    = {Phuc Nguyen and
               Ikuya Yamada and
               Natthawut Kertkeidkachorn and
               Ryutaro Ichise and
               Hideaki Takeda},
  title     = {SemTab 2021: Tabular Data Annotation with MTab Tool},
  booktitle = {SemTab@ISWC 2021},
  series    = {{CEUR} Workshop Proceedings},
  volume    = {3103},
  pages     = {92--101},
  publisher = {CEUR-WS.org},
  year      = {2021},
  url       = {http://ceur-ws.org/Vol-3103/paper8.pdf},
}

Contact

Phuc Nguyen ([email protected])

mtab_tool's Issues

The API does not work now

Hi~
I want to build a system related to using an external knowledge base to provide useful information for end-users. I found your API is suitable for my project and I want to cite your work. However, the API of search entity does not work now. I try to ask for help and figure out the reason.

——————Thank you. I notice that you are carrying out maintenance!

Return all entity candidates and their confidence scores as the final CEA annotation results

I got a new request to enable returning all entity candidates and their confidence scores as the final results of the CEA task.

The user could find the example as turning debug=True in the request.

Please change the dir_table to your table location.

mtab_tool/api/annotator/m_table_annotation.py

Line 111 in 66ec28f

debug=True, # return all candidates, and their confidence scores in CEA tasks

The annotation results will look like this file.
mtab_annotation_result.txt

Replicate MTab results on Tough Table dataset

I got a request to replicate MTab results on the Tough Table dataset with MTab API.

The results can be replicated with this script: https://github.com/phucty/mtab_tool/blob/master/run_2t.py
Other resources:
2T dataset: https://github.com/phucty/mtab_tool/blob/master/data/semtab/tables.zip
Annotation Results: https://github.com/phucty/mtab_tool/blob/master/results.zip

The final result is very close to our results in SemTab 2020. (F1: 0.895 vs. 0.907 as reported in our SemTab 2020 report)
Setting:

- search_mode="a", (Using aggregation search: Fuzzy + keyword )
- search_limit=100 (Get top 100 relevant entity candidates),
- search_expensive=True (Do not stop early once got a good candidate),
- chunk_size=200 (Split big table to smaller (200 line/table)),

Result:

180/180. Z4M8AT89: 100%|██████████████████████████████████████| 667244/667244 [21:53:57<00:00,  8.46it/s]
{
    "ALL": {
        "precision": 0.8954061559997541,
        "recall": 0.8953726073220591,
        "f1": 0.8953893813466541,
        "correct": 597432,
        "gt": 667244,
        "submit": 667219
    }
}

The differences in the results could be explained as:

1. Wikidata change over time: The online API of the MTab system is built on a Wikidata dump in 2021, while the 2T dataset results in 2020. There are many changes in Wikidata, so we have a difference in the final results.
2. Since we have limited resources to host MTab services, we have to split big tables into chunks and feed them to MTab API, so it could also affect the final results.

API connection refused

Hello~ The API worked fine yesterday, but today the connection was rejected.
I annotated a table with curl -X POST -F file=@"0.t0.csv" https://mtab.app/api/v1/mtab, but got curl: (7) Failed to connect to mtab.app port 443: Connection refused. And inference can't be opened, too.

Are you going to release the frameworks code?