Practical overview
When it fits
Use Text2Keywords when you need a Russian-language keyword-generation model that can be loaded in a Python workflow, or when you want to explore adapting T5 with paired text and keyword examples.
Who it is for
- Developers building Russian-language text-processing or content-tagging workflows.
- Practitioners who want to explore keyword generation with pretrained T5 models.
- Teams preparing paired text-and-keyword examples for model customization.
Documented uses
- Generate a keyword list from a Russian article or other text.
- Add Russian keyword extraction to a content-processing pipeline.
- Prepare a custom CSV dataset of input text and desired keyword output for training.
Requirements & limits
The repository is public and specifies an MIT license. Its README documents loading the pretrained model weights and provides links to inference and training notebooks.
- The README identifies Russian as the supported language.
- Both keyT5-base and keyT5-large weights are listed; no comparative performance guidance is provided.
- The example uses Hugging Face Transformers, SentencePiece and PyTorch.
- The README warns that generated keyword lists may contain duplicates.
- Custom training is described using CSV input and target columns and a linked notebook.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Generate keywords from Russian text with a pretrained model
Load keyT5-base or keyT5-large and generate a keyword list from an input text.
Before you start
- Python environment
- Install Transformers and SentencePiece using the repository’s documented command
- PyTorch, used by the inference example
- A Russian-language input text
Steps
Install the documented dependencies: transformers and sentencepiece.
Load the tokenizer and T5ForConditionalGeneration weights from the keyT5-base or keyT5-large model identifier using Hugging Face Transformers.
Tokenize the input text and run model generation with PyTorch, following the repository’s example.
Decode the generated text and split it into the keyword list, as shown in the example.
Expected outcome
A list of generated Russian keywords for the supplied text.
Guide limitations
- The repository says duplicate keywords may occur; its sample removes adjacent repeated entries but still shows repeated terms in the result.
- The repository identifies Russian as the supported language.
Prepare a CSV for custom keyword-model training
Format input texts and desired keyword outputs for the repository’s linked training workflow.
Before you start
- A CSV dataset with input text and target keyword output arranged as the repository’s X and Y example
- Access to the linked training notebook
Steps
Create CSV rows with source text in the input column and the desired keyword text in the output column.
Review the repository’s linked training notebook for the training stage.
Expected outcome
A dataset in the format described by the repository and a documented next step for model training.
Guide limitations
- The README links to the training notebook but the available instructions here do not detail its full execution procedure.
- No results are established for a particular custom dataset.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 18 October 2026
- Original website title
- 0x7o/text2keywords · Official homepage title · retained in its source language