|
---
|
|
license: apache-2.0
|
|
task_categories:
|
|
- question-answering
|
|
language:
|
|
- th
|
|
- en
|
|
tags:
|
|
- art
|
|
pretty_name: Lexitron 2.0 Prompt Finetuning Dataset
|
|
size_categories:
|
|
- 100K<n<1M
|
|
maintainer: Kobkrit Viriyayudhakorn (kobkrit@iapp.co.th)
|
|
---
|
|
# Lexitron 2.0 Prompt Finetuning Dataset
|
|
|
|
This dataset is derived from Lexitron 2.0, a Thai-English dictionary developed by NECTEC. It has been processed and formatted for prompt finetuning tasks.
|
|
|
|
## Dataset Description
|
|
|
|
The dataset consists of two main files:
|
|
|
|
1. `lexitron2_telex_finetune.qwen2.txt` - Thai to English lexicon entries in Qwen2 conversation format
|
|
2. `lexitron2_telex_finetune.jsonl` - Thai to English lexicon entries in JSONL format
|
|
|
|
Each file contains dictionary entries formatted in their respective formats.
|
|
|
|
## Syntax and Example Dataset
|
|
|
|
### Syntax
|
|
|
|
The dataset files follow specific formats for ease of use in prompt finetuning tasks.
|
|
|
|
#### `lexitron2_telex_finetune.jsonl`
|
|
|
|
Each entry in this file follows the generic prompt format:
|
|
|
|
```json
|
|
{
|
|
"instruction": "<instruction in English or Thai>",
|
|
"output": "<translation or definition in Thai>"
|
|
}
|
|
```
|
|
|
|
- **instruction**: A string that provides the translation task or query. It can be in English or Thai.
|
|
- **output**: A string that provides the translation or definition in Thai.
|
|
|
|
#### `lexitron2_telex_finetune.qwen2.txt`
|
|
|
|
Each entry in this file follows the Qwen2 conversation format:
|
|
|
|
```text
|
|
<|im_start|>system\n
|
|
คุณคือผู้ช่วยตอบคำถามที่ฉลาดและซื่อสัตย์<|im_end|>\n
|
|
<|im_start|>user\n
|
|
คำว่า 'example_word' แปลว่าอะไร<|im_end|>\n
|
|
<|im_start|>assistant\n
|
|
คำว่า 'example_word' เป็นคำประเภท Noun (คำนาม) แปลเป็นภาษาอังกฤษว่า example\n
|
|
ความหมาย: ตัวอย่างความหมาย\n
|
|
ตัวอย่างประโยค: ตัวอย่างการใช้ในประโยค<|im_end|>
|
|
```
|
|
|
|
The Qwen2 format consists of three main components:
|
|
- **system**: Defines the AI assistant's role and characteristics
|
|
- **user**: Contains the translation query or question (in Thai or English)
|
|
- **assistant**: Provides the structured response including:
|
|
- Word type/category
|
|
- English translation
|
|
- Definition
|
|
- Example sentences (when available)
|
|
- Synonyms (when available)
|
|
- Antonyms (when available)
|
|
- Related words (when available)
|
|
- Classifiers (when available)
|
|
|
|
## Example Dataset
|
|
|
|
Here is an example of how the dataset is structured:
|
|
```json
|
|
{"instruction": "Please translate 'a' into Thai", "output": "คำว่า 'a' เป็นคำประเภท Determiner แปลเป็นภาษาไทยว่า 'หนึ่ง (คำนำหน้าคำนามเพื่อแสดงว่าคำนามนั้นๆ ไม่ชี้เฉพาะ)'"}
|
|
```
|
|
|
|
```text
|
|
<|im_start|>system\nคุณคือผู้ช่วยตอบคำถามที่ฉลาดและซื่อสัตย์<|im_end|>\n<|im_start|>user\nคำว่า 'ดังกล่าวข้างต้น' สามารถใช้ในประโยคอย่างไรได้บ้าง<|im_end|>\n<|im_start|>assistant\nคำว่า 'ดังกล่าวข้างต้น' เป็นคำประเภท Pronoun (คำสรรพนาม) แปลเป็นภาษาอังกฤษว่า abovementioned\nคำที่มีความหมายเหมือนกัน: ดังกล่าว\nตัวอย่างประโยค: หน่วยงานของเราสามารถรับบทบาทได้เป็นอย่างดี ตามสภาพความพร้อมด้านต่างๆ ดังกล่าวข้างต้น<|im_end|>
|
|
```
|
|
|
|
|
|
|