# Missing Word Position in Kazakh Sentences: Data Description

*English translation by SOTA – AI Community of the Russian original. Licensed CC BY-NC-SA 4.0, like the original. Organisers who would like this translation removed can email sota.ai.community@gmail.com.*

*Contest 4 (29 June 2025) of the Kazakhstan IOAI Team Selection Test (Отборочные на IOAI), 2025. This is the Data tab ("Dataset Description") of the Kaggle competition "TST Day4 Upsolving". Original: [kaggle.com/competitions/tst-day-4-upsolving/data](https://www.kaggle.com/competitions/tst-day-4-upsolving/data). The Data tab lists 3 files (7.07 MB): `train.csv`, `public_test.csv` and `sample_submission.csv`; the description below does not name them.*

**Training dataset**
The training set was built from digitised and optically recognised (OCR) scans of school textbooks in the Kazakh language. Because the source materials are scanned pages, the dataset contains noise and errors. Typical problems:

* incorrectly recognised characters and punctuation marks
* fragmented or inconsistent sentence structures
* random inclusions of interface elements and annotation artefacts

Nevertheless, the set covers a variety of syntactic constructions and lexical expressions typical of educational texts. It is intended to model real data-preprocessing problems for low-resource languages and archival scanned materials. Participants are invited to build models that are robust to noise and to partial corruption of the data, and that can generalise and work correctly even when recognition errors are present.

---

**Test dataset**
The test set was compiled from a separate corpus that was carefully cleaned of OCR artefacts. Each sentence contains exactly one missing word, and the participants' task is to predict the (zero-based) position of this word within the sentence. Data format:

* `ID`
* `masked_sentence` — the sentence, with the missing word replaced by a special token

The test data contain no noise artefacts and fully reflect correct Kazakh text, which makes an objective assessment of model quality possible.