| --- |
| language: |
| - kk |
| license: cc-by-nc-4.0 |
| task_categories: |
| - text-generation |
| - question-answering |
| tags: |
| - kazakh |
| - agentic-ai |
| - tool-use |
| - function-calling |
| - tool-selection |
| - disambiguation |
| - web-search |
| - news-search |
| - instruction-following |
| - llm-agents |
| pretty_name: Kazakh Tool Selection and Disambiguation Dataset |
| size_categories: |
| - 1K<n<10K |
| --- |
| |
| # 🇰🇿 Kazakh Tool Selection and Disambiguation Dataset |
|
|
| ## Dataset Summary |
|
|
| **Kazakh Tool Selection and Disambiguation Dataset** is a Kazakh-language dataset designed for training and evaluating Large Language Models (LLMs) in agentic AI scenarios that require choosing the most appropriate tool from multiple available options. |
|
|
| The dataset focuses on tool-selection reasoning, where the assistant must understand the user’s intent, compare available tools, avoid unnecessary tool calls, and select the tool that best fits the task. It is especially useful for scenarios where several tools appear related but only one is appropriate, such as choosing between general web search and recent news search. |
|
|
|
|
| --- |
|
|
| ## 📊 Dataset Statistics |
|
|
| ### General Metrics |
|
|
| | Metric | Count | |
| | :--- | :--- | |
| | **Total Samples** | 3,579 | |
| | **Total Words** (approx.) | 1,375,613 | |
| | **Avg. Words per Sample** | 384 | |
|
|
| ### Word Count Distribution Per Field |
|
|
| The following table details the distribution of word counts across different fields in the dataset. |
|
|
| | Field | Mean | Median | Min | Max | Total Words | |
| | :--- | :--- | :--- | :--- | :--- | :--- | |
| | **answers** | 26.4 | 24.0 | 6 | 174 | 94,600 | |
| | **category** | 1.0 | 1.0 | 1 | 1 | 3,579 | |
| | **difficulty** | 1.0 | 1.0 | 1 | 1 | 3,579 | |
| | **id** | 1.0 | 1.0 | 1 | 1 | 3,579 | |
| | **query** | 14.8 | 14.0 | 3 | 42 | 52,923 | |
| | **tools** | 172.3 | 172.0 | 50 | 291 | 616,606 | |
| | **turns** | 167.9 | 165.0 | 64 | 502 | 600,747 | |
|
|
|
|
|  |
|
|
| --- |
|
|
| ## Dataset Structure |
|
|
| Each dataset instance represents a tool-selection scenario. A sample usually includes a Kazakh user query, several available tools, the expected tool call, a simulated tool response, and the final assistant answer. |
|
|
| ### Data Fields |
|
|
| - **`id`**: A unique identifier for the sample. |
|
|
| - **`category`**: The task category. For example, `02_tool_selection_disambiguation` indicates that the sample focuses on selecting the correct tool among multiple possible tools. |
|
|
| - **`query`**: The original user request in Kazakh. Some queries may include mixed-language terms, such as English technical words inside Kazakh sentences. |
|
|
| - **`tools`**: A list of available tools. Each tool contains: |
| - `name`: the tool name; |
| - `description`: the tool’s purpose; |
| - `parameters`: the expected argument schema. |
|
|
| - **`difficulty`**: The difficulty level of the sample. |
|
|
| - **`answers`**: The expected tool call or calls. This field usually contains the selected tool name and serialized JSON arguments. |
|
|
| - **`turns`**: The full interaction trace, including: |
| - user request; |
| - assistant planning or tool-selection reasoning; |
| - assistant tool call; |
| - mock tool response; |
| - final answer grounded in the tool output. |
|
|
| --- |
|
|
| ## Data Instance |
|
|
| Below is one representative example from the dataset. |
|
|
| ```json |
| { |
| "id": "asem_kk_02_tool_selection_disambiguation_1751", |
| "category": "02_tool_selection_disambiguation", |
| "query": "Ең танымал programming тілдері қандай?", |
| "tools": [ |
| { |
| "name": "web.search", |
| "description": "Search the web for information", |
| "parameters": { |
| "query": { |
| "type": "string", |
| "description": "Search query", |
| "required": true |
| }, |
| "limit": { |
| "type": "int", |
| "description": "Number of results", |
| "required": false |
| } |
| } |
| }, |
| { |
| "name": "news.search", |
| "description": "Search recent news articles", |
| "parameters": { |
| "query": { |
| "type": "string", |
| "description": "Search topic", |
| "required": true |
| }, |
| "language": { |
| "type": "string", |
| "description": "Language code", |
| "required": false |
| }, |
| "pageToken": { |
| "type": "string", |
| "description": "Pagination token", |
| "required": false |
| } |
| } |
| } |
| ], |
| "difficulty": "easy", |
| "answers": [ |
| { |
| "name": "web.search", |
| "arguments": "{\"query\": \"ең танымал программалау тілдері 2024\", \"limit\": 1}" |
| } |
| ], |
| "turns": [ |
| { |
| "role": "user", |
| "content": "Ең танымал programming тілдері қандай?" |
| }, |
| { |
| "role": "assistant", |
| "content": "web.search дерегін алу қажет, news.search дерегін алу қажет емес", |
| "meta": { |
| "plan": "Need web.search and don't need news.search", |
| "tags": [], |
| "difficulty": "hard" |
| } |
| }, |
| { |
| "role": "assistant", |
| "content": "", |
| "tool_call": { |
| "name": "web.search", |
| "arguments": "{\"query\": \"ең танымал программалау тілдері 2024\", \"limit\": 1}" |
| } |
| }, |
| { |
| "role": "tool", |
| "content": "{\"results\": [{\"title\": \"2024 жылы ең танымал программалау тілдері\", \"snippet\": \"2024 жылы ең танымал программалау тілдері қатарына Python, JavaScript, Java, және C# кіреді.\", \"url\": \"https://www.example.com/top-programming-languages-2024\"}]}" |
| }, |
| { |
| "role": "assistant", |
| "content": "2024 жылы ең танымал программалау тілдері ретінде Python, JavaScript, Java және C# көрсетілген. (Дерек көзі: web.search)" |
| } |
| ] |
| } |
| ``` |
|
|
| ## Funding |
|
|
| This dataset was developed as part of the project funded by the Ministry of Science and Higher Education of the Republic of Kazakhstan under Grant No. BR24993001, “Creation of a Large Language Model (LLM) to Support the Kazakh Language and Advance Technological Development.” |
|
|
| ## Citation |
|
|
| If you use this dataset in your research, please cite the following article: |
|
|
| ### APA |
|
|
| Kadyrbek, N., Tuimebayev, Z., Mansurova, M., & Viegas, V. (2025). The development of small-scale language models for low-resource languages, with a focus on Kazakh and direct preference optimization. *Big Data and Cognitive Computing, 9*(5), 137. [https://doi.org/10.3390/bdcc9050137](https://doi.org/10.3390/bdcc9050137) |
|
|
| ### BibTeX |
|
|
| ```bibtex |
| @article{kadyrbek2025development, |
| title = {The Development of Small-Scale Language Models for Low-Resource Languages, with a Focus on Kazakh and Direct Preference Optimization}, |
| author = {Kadyrbek, Nurgali and Tuimebayev, Zhanseit and Mansurova, Madina and Viegas, Vitor}, |
| journal = {Big Data and Cognitive Computing}, |
| volume = {9}, |
| number = {5}, |
| pages = {137}, |
| year = {2025}, |
| publisher = {MDPI}, |
| doi = {10.3390/bdcc9050137}, |
| url = {https://www.mdpi.com/2504-2289/9/5/137} |
| } |
| ``` |
|
|
