rut5-base-summ-dialogsum Open-source Russian Dialogue Summarization Model - Precisely Extract the Core Content of Conversations

Home

Rut5 Base Summ Dialogsum

Developed by Kekega

A Russian dialogue summarization model fine-tuned on d0rj/rut5-base-summ, trained on the DialogSum dataset

Text Generation

Transformers

#Russian Text Summarization #Dialogue Summarization Optimization #Rouge Metric Optimization

Downloads 15

Release Time : 12/20/2023

Model Overview

This model is a T5 variant specifically designed for Russian dialogue summarization tasks, capable of extracting key information from dialogue texts to generate concise summaries

Model Features

Russian Dialogue Summarization Optimization

Fine-tuned specifically for Russian dialogue scenarios, better understanding dialogue context

Based on T5 Architecture

Utilizes Transformer architecture with powerful sequence-to-sequence conversion capabilities

Multi-Round Training Optimization

Refined through 25 rounds of fine-tuning, with stable Rouge metric performance

Model Capabilities

Dialogue Text Summarization

Key Information Extraction

Russian Text Generation

Use Cases

Customer Service Dialogue Analysis

Customer Service Dialogue Summary

Automatically generates summaries of key issues and solutions from customer service dialogues

Rouge1 score 33.51

Meeting Minutes

Meeting Key Points Extraction

Extracts decision points and action items from Russian meeting records

🚀 rut5-base-summ-dialogsum

This model is a fine - tuned version of d0rj/rut5-base-summ on the None dataset. It offers a set of evaluation metrics that can be used to assess its performance.

📚 Documentation

Model Information

Property	Details
Base Model	d0rj/rut5-base-summ
Tags	generated_from_trainer
Metrics	rouge

Evaluation Results

This model achieves the following results on the evaluation set:

Loss: 1.1263
Rouge1: 33.5111
Rouge2: 0.1696
Rougel: 33.4559
Rougelsum: 33.4934
Gen Len: 4.1546

Training Procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 5e - 05
train_batch_size: 8
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e - 08
lr_scheduler_type: linear
num_epochs: 25

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
2.0946	1.0	786	1.7462	45.4252	0.0	45.4009	45.4139	4.0464
1.7182	2.0	1572	1.5005	44.9295	0.0	44.9183	44.9108	4.1126
1.5304	3.0	2358	1.3826	39.5888	0.0	39.5811	39.5646	4.1698
1.4261	4.0	3144	1.3121	30.1735	0.0	30.1127	30.1415	4.1520
1.3252	5.0	3930	1.2641	35.7738	0.0	35.7408	35.7858	3.8791
1.2878	6.0	4716	1.2353	33.0773	0.0	32.9682	33.0551	3.7252
1.2068	7.0	5502	1.2051	34.4094	0.0	34.3902	34.3884	3.7729
1.1763	8.0	6288	1.1952	33.0914	0.1908	33.0267	33.0472	3.9739
1.1346	9.0	7074	1.1798	33.9606	0.0	33.9335	33.979	4.1768
1.1044	10.0	7860	1.1632	32.9529	0.0	32.9367	32.9396	4.1673
1.1073	11.0	8646	1.1499	34.0904	0.0	34.0659	34.1317	4.1934
1.0619	12.0	9432	1.1516	32.9502	0.0	32.9056	32.9376	4.0312
1.0365	13.0	10218	1.1478	31.68	0.0	31.6488	31.7003	4.0293
1.0161	14.0	11004	1.1427	32.6651	0.0424	32.6345	32.6538	4.1113
0.9805	15.0	11790	1.1343	34.0304	0.0636	33.9433	33.999	4.0674
0.9661	16.0	12576	1.1309	34.8704	0.0848	34.8014	34.8501	4.0681
0.9511	17.0	13362	1.1348	32.8744	0.0	32.8277	32.8547	4.1081
0.9392	18.0	14148	1.1326	32.9349	0.1908	32.8895	32.9376	4.2627
0.9341	19.0	14934	1.1263	33.5111	0.1696	33.4559	33.4934	4.1546
0.9396	20.0	15720	1.1349	33.9121	0.2545	33.8438	33.8993	4.1705
0.9314	21.0	16506	1.1276	33.0779	0.106	33.0546	33.0903	4.1399
0.8987	22.0	17292	1.1333	33.8566	0.1696	33.7943	33.843	4.1419
0.8895	23.0	18078	1.1343	33.6108	0.1484	33.5738	33.636	4.2328
0.8847	24.0	18864	1.1355	33.4257	0.2757	33.3804	33.4495	4.1711
0.8832	25.0	19650	1.1355	33.6211	0.3393	33.5937	33.636	4.1959

Framework versions

Transformers 4.35.2
Pytorch 2.0.1+cu117
Datasets 2.15.0
Tokenizers 0.15.0

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご