Instructor Base

由 hkunlp 开发

基于T5架构的文本嵌入模型，专注于句子相似度计算和文本检索任务，在多个基准测试中表现优异。

文本嵌入

Transformers

英语

开源协议:Apache-2.0 #文本语义相似度 #多任务文本嵌入 #英语信息检索

下载量 13.22k

发布时间 : 12/20/2022

模型介绍

内容详情

替代品

模型简介

该模型是一个基于T5架构的文本嵌入模型，主要用于生成高质量的句子嵌入向量，支持信息检索、文本分类、聚类和语义相似度计算等多种自然语言处理任务。

模型特点

多任务性能优异

在MTEB基准测试的多个任务中表现优秀，包括分类、聚类和检索任务

高效文本嵌入

能够生成高质量的句子嵌入向量，适用于大规模信息检索场景

广泛适用性

支持多种下游NLP任务，包括相似度计算、分类和聚类等

模型能力

句子相似度计算

文本嵌入生成

信息检索

文本分类

文本聚类

语义搜索

文本重排序

使用案例

电子商务

产品评论分类

对亚马逊产品评论进行情感分析分类

在AmazonPolarity分类任务中达到88.36%准确率

反事实检测

识别亚马逊产品评论中的反事实陈述

在AmazonCounterfactual分类任务中达到86.21%准确率

金融

银行客服分类

对银行客户咨询进行分类

在Banking77分类任务中达到77.04%准确率

学术研究

论文聚类

对arXiv和biorxiv论文进行主题聚类

在ArxivClusteringP2P任务中达到39.68 v_measure分数

pipeline_tag: 句子相似度 tags:

文本嵌入
嵌入向量
信息检索
beir
文本分类
语言模型
文本聚类
文本语义相似度
文本评估
提示检索
文本重排序
sentence-transformers
特征提取
句子相似度
transformers
t5
英语
句子相似度
natural_questions
ms_marco
fever
hotpot_qa
mteb language: en inference: false license: apache-2.0 model-index:
name: final_base_results results:
- task: type: 分类 dataset: type: mteb/amazon_counterfactual name: MTEB AmazonCounterfactualClassification (en) config: en split: test revision: e8379541af4e31359cca9fbcf4b00f2671dba205 metrics:
  - type: 准确率 value: 86.2089552238806
  - type: 平均精度 value: 55.76273850794966
  - type: f1分数 value: 81.26104211414781
- task: type: 分类 dataset: type: mteb/amazon_polarity name: MTEB AmazonPolarityClassification config: default split: test revision: e2d317d38cd51312af73b3d32a06d1a08b442046 metrics:
  - type: 准确率 value: 88.35995000000001
  - type: 平均精度 value: 84.18839957309655
  - type: f1分数 value: 88.317619250081
- task: type: 分类 dataset: type: mteb/amazon_reviews_multi name: MTEB AmazonReviewsClassification (en) config: en split: test revision: 1399c76144fd37290681b995c656ef9b2e06e26d metrics:
  - type: 准确率 value: 44.64
  - type: f1分数 value: 42.48663956478136
- task: type: 检索 dataset: type: arguana name: MTEB ArguAna config: default split: test revision: None metrics:
  - type: map@1 value: 27.383000000000003
  - type: map@10 value: 43.024
  - type: map@100 value: 44.023
  - type: map@1000 value: 44.025999999999996
  - type: map@3 value: 37.684
  - type: map@5 value: 40.884
  - type: mrr@1 value: 28.094
  - type: mrr@10 value: 43.315
  - type: mrr@100 value: 44.313
  - type: mrr@1000 value: 44.317
  - type: mrr@3 value: 37.862
  - type: mrr@5 value: 41.155
  - type: ndcg@1 value: 27.383000000000003
  - type: ndcg@10 value: 52.032000000000004
  - type: ndcg@100 value: 56.19499999999999
  - type: ndcg@1000 value: 56.272
  - type: ndcg@3 value: 41.166000000000004
  - type: ndcg@5 value: 46.92
  - type: precision@1 value: 27.383000000000003
  - type: precision@10 value: 8.087
  - type: precision@100 value: 0.989
  - type: precision@1000 value: 0.099
  - type: precision@3 value: 17.093
  - type: precision@5 value: 13.044
  - type: recall@1 value: 27.383000000000003
  - type: recall@10 value: 80.868
  - type: recall@100 value: 98.86200000000001
  - type: recall@1000 value: 99.431
  - type: recall@3 value: 51.28
  - type: recall@5 value: 65.22
- task: type: 聚类 dataset: type: mteb/arxiv-clustering-p2p name: MTEB ArxivClusteringP2P config: default split: test revision: a122ad7f3f0291bf49cc6f4d32aa80929df69d5d metrics:
  - type: v_measure value: 39.68441054431849
- task: type: 聚类 dataset: type: mteb/arxiv-clustering-s2s name: MTEB ArxivClusteringS2S config: default split: test revision: f910caf1a6075f7329cdf8c1a6135696f37dbd53 metrics:
  - type: v_measure value: 29.188539728343844
- task: type: 重排序 dataset: type: mteb/askubuntudupquestions-reranking name: MTEB AskUbuntuDupQuestions config: default split: test revision: 2000358ca161889fa9c082cb41daa8dcfb161a54 metrics:
  - type: map value: 63.173362687519784
  - type: mrr value: 76.18860748362133
- task: type: STS dataset: type: mteb/biosses-sts name: MTEB BIOSSES config: default split: test revision: d3fb88f8f02e40887cd149695127462bbcf29b4a metrics:
  - type: 余弦相似度斯皮尔曼 value: 82.30789953771232
- task: type: 分类 dataset: type: mteb/banking77 name: MTEB Banking77Classification config: default split: test revision: 0fd18e25b25c072e09e0d92ab615fda904d66300 metrics:
  - type: 准确率 value: 77.03571428571428
  - type: f1分数 value: 75.87384305045917
- task: type: 聚类 dataset: type: mteb/biorxiv-clustering-p2p name: MTEB BiorxivClusteringP2P config: default split: test revision: 65b79d1d13f80053f67aca9498d9402c2d9f1f40 metrics:
  - type: v_measure value: 32.98041170516364
- task: type: 聚类 dataset: type: mteb/biorxiv-clustering-s2s name: MTEB BiorxivClusteringS2S config: default split: test revision: 258694dd0231531bc1fd9de6ceb52a0853c6d908 metrics:
  - type: v_measure value: 25.71652988451154
- task: type: 检索 dataset: type: BeIR/cqadupstack name: MTEB CQADupstackAndroidRetrieval config: default split: test revision: None metrics:
  - type: map@1 value: 33.739999999999995
  - type: map@10 value: 46.197
  - type: map@100 value: 47.814
  - type: map@1000 value: 47.934
  - type: map@3 value: 43.091
  - type: map@5 value: 44.81
  - type: mrr@1 value: 41.059
  - type: mrr@10 value: 52.292
  - type: mrr@100 value: 52.978
  - type: mrr@1000 value: 53.015
  - type: mrr@3 value: 49.976
  - type: mrr@5 value: 51.449999999999996
  - type: ndcg@1 value: 41.059
  - type: ndcg@10 value: 52.608
  - type: ndcg@100 value: 57.965
  - type: ndcg@1000 value: 59.775999999999996
  - type: ndcg@3 value: 48.473
  - type: ndcg@5 value: 50.407999999999994
  - type: precision@1 value: 41.059
  - type: precision@10 value: 9.943
  - type: precision@100 value: 1.6070000000000002
  - type: precision@1000 value: 0.20500000000000002
  - type: precision@3 value: 23.413999999999998
  - type: precision@5 value: 16.481
  - type: recall@1 value: 33.739999999999995
  - type: recall@10 value: 63.888999999999996
  - type: recall@100 value: 85.832
  - type: recall@1000 value: 97.475
  - type: recall@3 value: 51.953
  - type: recall@5 value: 57.498000000000005
- task: type: 检索 dataset: type: BeIR/cqadupstack name: MTEB CQADupstackEnglishRetrieval config: default split: test revision: None metrics:
  - type: map@1 value: 31.169999999999998
  - type: map@10 value: 41.455
  - type: map@100 value: 42.716
  - type: map@1000 value: 42.847
  - type: map@3 value: 38.568999999999996
  - type: map@5 value: 40.099000000000004
  - type: mrr@1 value: 39.427
  - type: mrr@10 value: 47.818
  - type: mrr@100 value: 48.519
  - type: mrr@1000 value: 48.558
  - type: mrr@3 value: 45.86
  - type: mrr@5 value: 46.936
  - type: ndcg@1 value: 39.427
  - type: ndcg@10 value: 47.181
  - type: ndcg@100 value: 51.737
  - type: ndcg@1000 value: 53.74
  - type: ndcg@3 value: 43.261
  - type: ndcg@5 value: 44.891
  - type: precision@1 value: 39.427
  - type: precision@10 value: 8.847
  - type: precision@100 value: 1.425
  - type: precision@1000 value: 0.189
  - type: precision@3 value: 20.785999999999998
  - type: precision@5 value: 14.560999999999998
  - type: recall@1 value: 31.169999999999998
  - type: recall@10 value: 56.971000000000004
  - type: recall@100 value: