AI Challenger 2018情感分析赛道资料汇总

男娘i 2022-02-20 05:17 235阅读 0赞

## 汇总 ##

AI Challenger 2018 已近尾声，各赛道top选手已经结束了代码核验，正在准备12月18、19日 AI Challenger 决赛答辩材料的路上。在本年度 AI Challenger 即将尘埃落定之时，这里整理一批目前网上可见的文本挖掘相关赛道的解决方案和代码，欢迎补充，同时感谢github，感谢各位开源的同学。

**[细粒度用户评论情感分析][Link 1]**

> 在线评论的细粒度情感分析对于深刻理解商家和用户、挖掘用户情感等方面有至关重要的价值，并且在互联网行业有极其广泛的应用，主要用于个性化推荐、智能搜索、产品反馈、业务安全等。本次比赛我们提供了一个高质量的海量数据集，共包含6大类20个细粒度要素的情感倾向。参赛人员需根据标注的细粒度要素的情感倾向建立算法，对用户评论进行情感挖掘，组委将通过计算参赛者提交预测值和场景真实值之间的误差确定预测正确率，评估所提交的预测算法。

貌似是最火爆的一个赛道，Testa 提交队伍有468支，详细介绍请参考该赛道主页：[https://challenger.ai/competition/fsauor2018][Link 1]

相关代码或解决方案：

TOP1. 冠军解决方案：  
1） [AI Challenger 2018 冠军 PPT 分享---细粒度情感分析赛道][AI Challenger 2018 _ PPT _---]  
2） 代码：[https://github.com/chenghuige/wenzheng/tree/master/projects/ai2018/sentiment][https_github.com_chenghuige_wenzheng_tree_master_projects_ai2018_sentiment]  
阁子大神补充：开源了所有代码但是我最近没有时间整理流程 估计复现会有困难 仅供参考；我这个比较乱如果看的话 参考一下 algos tf模型 torch-algos pyt模型就可以了 fast elmo参考 torch-lm-train.py 我这边都是先生成tfrecord 那部分代码在prepare 需要先转简体 再分好词 整体复现比较麻烦 ；python path需要设置 下载路径utils 这样能找到下面的melt等路径

TOP2. 亚军解决方案：  
[AI Challenger2018情感分析赛道亚军PPT分享][AI Challenger2018_PPT]

TOP4. 决赛第4名解决方案：  
[AI Challenger 2018 第4名PPT分享---细粒度情感分析赛道][AI Challenger 2018 _4_PPT_---]

1. 官方baseline，基于SVM: [sentiment\_analysis2018\_baseline][sentiment_analysis2018_baseline]  
[https://github.com/AIChallenger/AI\_Challenger\_2018/tree/master/Baselines/sentiment\_analysis2018\_baseline][sentiment_analysis2018_baseline]

2. 基于fastText的baseline: [AI Challenger 2018 Sentiment Analysis Baseline with fastText][]  
2.1 [https://github.com/panyang/fastText-for-AI-Challenger-Sentiment-Analysis][AI Challenger 2018 Sentiment Analysis Baseline with fastText]  
2.2 文章：[AI Challenger 2018 细粒度用户评论情感分析 fastText Baseline][AI Challenger 2018 _ fastText Baseline]

3. 基于 SVM 的细粒度情感分析: [https://github.com/scruel/sentiment\_analysis][https_github.com_scruel_sentiment_analysis]

4. 第16名解决方案: [https://github.com/xueyouluo/fsauor2018][https_github.com_xueyouluo_fsauor2018]

5. 第17名解决方案：[https://github.com/BigHeartC/Al\_challenger\_2018\_sentiment\_analysis][https_github.com_BigHeartC_Al_challenger_2018_sentiment_analysis]

6. 基于Bert的尝试：[https://github.com/brightmart/sentiment\_analysis\_fine\_grain][https_github.com_brightmart_sentiment_analysis_fine_grain]

7. ai challenger Competitions 1: Fine-grained Sentiment Analysis of User Reviews：  
[https://github.com/ShawnXiha/Fine-grained-Sentiment-Analysis-of-User-Reviews][https_github.com_ShawnXiha_Fine-grained-Sentiment-Analysis-of-User-Reviews]

8. 细粒度用户评论情感分析 (0.70201)：[https://github.com/pengshuang/AI-Comp][https_github.com_pengshuang_AI-Comp]  
8.1 相关文章1：[https://zhuanlan.zhihu.com/p/47207009][https_zhuanlan.zhihu.com_p_47207009]  
8.2 相关文章2：[https://zhuanlan.zhihu.com/p/47278559][https_zhuanlan.zhihu.com_p_47278559]

9. [AI Challenger 细粒度用户评论情感分析线上0.62][AI Challenger _0.62]: [https://zhuanlan.zhihu.com/p/44857751][AI Challenger _0.62]

## 1.AI Challenger 2018 冠军 PPT 分享---细粒度情感分析赛道 ##

PPT内容如下：文末打包下载

![78dcc462-7572-47bf-9fbf-675f7ffedd08][]

![4852aac3-0f92-4903-a946-5f1df4e6dfc6][]

![92238d8d-58df-4db4-bf7e-e699db692ccb][]

![07f4e4a3-7b58-4bcd-9daa-3e136035b6dd][]

![80c761dd-90e4-4a8f-b7b3-8c0c951f5bb3][]

![fcc37f0c-0caf-4fdc-ad92-27d2c5c90757][]

![4f3a5d6e-5eb6-4837-b10d-7b11c60ec479][]

![8f3df6b4-5a41-4375-87dd-139bf87361ba][]

![5b12dd79-d105-49e2-9b46-3f77a3db57d0][]

![14614c5b-30bb-4467-b550-2ca8dc6d3b1c][]

![0cb05f59-c373-42a8-ab87-42bc67f89b63][]

![5a256a3b-1504-43b6-8506-4ee0f0e2a48b][]

![7464fe69-df13-4a9c-81e2-9cd4553d30e2][]

![2dab154e-ea78-4c1f-a1f7-b5382bb267ea][]

![02f928e2-bc1b-4f35-b884-1b6552ae8fa6][]

![2880a23c-3e27-4edc-b5a4-e72aaf4a6ecd][]

![394669ce-4cbf-44dd-b61c-153fc163b384][]

![b8b4b349-bab4-4d74-a9d8-8233032c9f44][]

![a328e452-42b6-4c4b-b6d0-638ddd4be992][]

![80895bbc-97e9-4579-a651-a6e386c85605][]

![769b5b1c-140c-4676-a43a-a1a4f67f48fe][]

![bc473cc8-46c7-4049-895e-9cb293494c83][]

![ef7b6b86-9bc1-4ab9-82a2-014d843a608e][]

![0d712459-001c-470b-9f98-0184e901deb3][]

![e426a9f0-d83a-4b74-8f3c-708889b41472][]

![f85cf8ef-cd33-4971-b8d3-c9edd4267423][]

![18fa852d-5123-4568-9955-32e1b4ad9d37][]

![d090936f-45f7-405f-b183-f77d61cfd9aa][]

![6eecce13-29a1-47f7-b617-93bc66c80038][]

![9d8755b8-e070-4a41-b536-e1314e76d153][]

![cc861c84-e464-4a2f-8e38-90e3114f0ce1][]

![637196d4-db08-4adf-887a-d0d5b2d5e80f][]

![f0d393e6-3582-4444-a6bb-ea99e21703ad][]

![05e27d23-d06a-4bf7-b9d1-f29bd55d6e75][]

昨天分享了阁子大神的PPT：[AI Challenger 2018 冠军 PPT 分享---细粒度情感分析赛道][AI Challenger 2018 _ PPT _--- 1] ，今天他就在 AI Challenger 细粒度比赛群里公布了的代码，不过有几句补充：

开源了所有代码但是我最近没有时间整理流程 估计复现会有困难 仅供参考；我这个比较乱如果看的话 参考一下 algos tf模型 torch-algos pyt模型就可以了 fast elmo参考 torch-lm-train.py 我这边都是先生成tfrecord 那部分代码在prepare 需要先转简体 再分好词 整体复现比较麻烦 ；python path需要设置 下载路径utils 这样能找到下面的melt等路径

Github链接：https://github.com/chenghuige/wenzheng/tree/master/projects/ai2018/sentiment

点击“阅读原文”可直达，有心的同学赶紧学习吧，这可是价值千金的实战学习资料，感谢阁子无私的分享。

# #

[Link 1]: https://challenger.ai/competition/fsauor2018
[AI Challenger 2018 _ PPT _---]: https://mp.weixin.qq.com/s/W0PhbE8149nD3Venmy33tw
[https_github.com_chenghuige_wenzheng_tree_master_projects_ai2018_sentiment]: https://github.com/chenghuige/wenzheng/tree/master/projects/ai2018/sentiment
[AI Challenger2018_PPT]: https://mp.weixin.qq.com/s/SycD5rGNeK5NwYwfmgCKoA
[AI Challenger 2018 _4_PPT_---]: https://mp.weixin.qq.com/s/J6jPxIToPJsA7aSb7wzIuQ
[sentiment_analysis2018_baseline]: https://github.com/AIChallenger/AI_Challenger_2018/tree/master/Baselines/sentiment_analysis2018_baseline
[AI Challenger 2018 Sentiment Analysis Baseline with fastText]: https://github.com/panyang/fastText-for-AI-Challenger-Sentiment-Analysis
[AI Challenger 2018 _ fastText Baseline]: http://www.52nlp.cn/ai-challenger-2018-%E7%BB%86%E7%B2%92%E5%BA%A6%E7%94%A8%E6%88%B7%E8%AF%84%E8%AE%BA%E6%83%85%E6%84%9F%E5%88%86%E6%9E%90-fasttext-baseline
[https_github.com_scruel_sentiment_analysis]: https://github.com/scruel/sentiment_analysis
[https_github.com_xueyouluo_fsauor2018]: https://github.com/xueyouluo/fsauor2018
[https_github.com_BigHeartC_Al_challenger_2018_sentiment_analysis]: https://github.com/BigHeartC/Al_challenger_2018_sentiment_analysis
[https_github.com_brightmart_sentiment_analysis_fine_grain]: https://github.com/brightmart/sentiment_analysis_fine_grain
[https_github.com_ShawnXiha_Fine-grained-Sentiment-Analysis-of-User-Reviews]: https://github.com/ShawnXiha/Fine-grained-Sentiment-Analysis-of-User-Reviews
[https_github.com_pengshuang_AI-Comp]: https://github.com/pengshuang/AI-Comp
[https_zhuanlan.zhihu.com_p_47207009]: https://zhuanlan.zhihu.com/p/47207009
[https_zhuanlan.zhihu.com_p_47278559]: https://zhuanlan.zhihu.com/p/47278559
[AI Challenger _0.62]: https://zhuanlan.zhihu.com/p/44857751
[78dcc462-7572-47bf-9fbf-675f7ffedd08]: https://file.ai100.com.cn/files/sogou-articles/original/78dcc462-7572-47bf-9fbf-675f7ffedd08/78dcc462-7572-47bf-9fbf-675f7ffedd08
[4852aac3-0f92-4903-a946-5f1df4e6dfc6]: https://file.ai100.com.cn/files/sogou-articles/original/4852aac3-0f92-4903-a946-5f1df4e6dfc6/4852aac3-0f92-4903-a946-5f1df4e6dfc6
[92238d8d-58df-4db4-bf7e-e699db692ccb]: https://file.ai100.com.cn/files/sogou-articles/original/92238d8d-58df-4db4-bf7e-e699db692ccb/92238d8d-58df-4db4-bf7e-e699db692ccb
[07f4e4a3-7b58-4bcd-9daa-3e136035b6dd]: https://file.ai100.com.cn/files/sogou-articles/original/07f4e4a3-7b58-4bcd-9daa-3e136035b6dd/07f4e4a3-7b58-4bcd-9daa-3e136035b6dd
[80c761dd-90e4-4a8f-b7b3-8c0c951f5bb3]: https://file.ai100.com.cn/files/sogou-articles/original/80c761dd-90e4-4a8f-b7b3-8c0c951f5bb3/80c761dd-90e4-4a8f-b7b3-8c0c951f5bb3
[fcc37f0c-0caf-4fdc-ad92-27d2c5c90757]: https://file.ai100.com.cn/files/sogou-articles/original/fcc37f0c-0caf-4fdc-ad92-27d2c5c90757/fcc37f0c-0caf-4fdc-ad92-27d2c5c90757
[4f3a5d6e-5eb6-4837-b10d-7b11c60ec479]: https://file.ai100.com.cn/files/sogou-articles/original/4f3a5d6e-5eb6-4837-b10d-7b11c60ec479/4f3a5d6e-5eb6-4837-b10d-7b11c60ec479
[8f3df6b4-5a41-4375-87dd-139bf87361ba]: https://file.ai100.com.cn/files/sogou-articles/original/8f3df6b4-5a41-4375-87dd-139bf87361ba/8f3df6b4-5a41-4375-87dd-139bf87361ba
[5b12dd79-d105-49e2-9b46-3f77a3db57d0]: https://file.ai100.com.cn/files/sogou-articles/original/5b12dd79-d105-49e2-9b46-3f77a3db57d0/5b12dd79-d105-49e2-9b46-3f77a3db57d0
[14614c5b-30bb-4467-b550-2ca8dc6d3b1c]: https://file.ai100.com.cn/files/sogou-articles/original/14614c5b-30bb-4467-b550-2ca8dc6d3b1c/14614c5b-30bb-4467-b550-2ca8dc6d3b1c
[0cb05f59-c373-42a8-ab87-42bc67f89b63]: https://file.ai100.com.cn/files/sogou-articles/original/0cb05f59-c373-42a8-ab87-42bc67f89b63/0cb05f59-c373-42a8-ab87-42bc67f89b63
[5a256a3b-1504-43b6-8506-4ee0f0e2a48b]: https://file.ai100.com.cn/files/sogou-articles/original/5a256a3b-1504-43b6-8506-4ee0f0e2a48b/5a256a3b-1504-43b6-8506-4ee0f0e2a48b
[7464fe69-df13-4a9c-81e2-9cd4553d30e2]: https://file.ai100.com.cn/files/sogou-articles/original/7464fe69-df13-4a9c-81e2-9cd4553d30e2/7464fe69-df13-4a9c-81e2-9cd4553d30e2
[2dab154e-ea78-4c1f-a1f7-b5382bb267ea]: https://file.ai100.com.cn/files/sogou-articles/original/2dab154e-ea78-4c1f-a1f7-b5382bb267ea/2dab154e-ea78-4c1f-a1f7-b5382bb267ea
[02f928e2-bc1b-4f35-b884-1b6552ae8fa6]: https://file.ai100.com.cn/files/sogou-articles/original/02f928e2-bc1b-4f35-b884-1b6552ae8fa6/02f928e2-bc1b-4f35-b884-1b6552ae8fa6
[2880a23c-3e27-4edc-b5a4-e72aaf4a6ecd]: https://file.ai100.com.cn/files/sogou-articles/original/2880a23c-3e27-4edc-b5a4-e72aaf4a6ecd/2880a23c-3e27-4edc-b5a4-e72aaf4a6ecd
[394669ce-4cbf-44dd-b61c-153fc163b384]: https://file.ai100.com.cn/files/sogou-articles/original/394669ce-4cbf-44dd-b61c-153fc163b384/394669ce-4cbf-44dd-b61c-153fc163b384
[b8b4b349-bab4-4d74-a9d8-8233032c9f44]: https://file.ai100.com.cn/files/sogou-articles/original/b8b4b349-bab4-4d74-a9d8-8233032c9f44/b8b4b349-bab4-4d74-a9d8-8233032c9f44
[a328e452-42b6-4c4b-b6d0-638ddd4be992]: https://file.ai100.com.cn/files/sogou-articles/original/a328e452-42b6-4c4b-b6d0-638ddd4be992/a328e452-42b6-4c4b-b6d0-638ddd4be992
[80895bbc-97e9-4579-a651-a6e386c85605]: https://file.ai100.com.cn/files/sogou-articles/original/80895bbc-97e9-4579-a651-a6e386c85605/80895bbc-97e9-4579-a651-a6e386c85605
[769b5b1c-140c-4676-a43a-a1a4f67f48fe]: https://file.ai100.com.cn/files/sogou-articles/original/769b5b1c-140c-4676-a43a-a1a4f67f48fe/769b5b1c-140c-4676-a43a-a1a4f67f48fe
[bc473cc8-46c7-4049-895e-9cb293494c83]: https://file.ai100.com.cn/files/sogou-articles/original/bc473cc8-46c7-4049-895e-9cb293494c83/bc473cc8-46c7-4049-895e-9cb293494c83
[ef7b6b86-9bc1-4ab9-82a2-014d843a608e]: https://file.ai100.com.cn/files/sogou-articles/original/ef7b6b86-9bc1-4ab9-82a2-014d843a608e/ef7b6b86-9bc1-4ab9-82a2-014d843a608e
[0d712459-001c-470b-9f98-0184e901deb3]: https://file.ai100.com.cn/files/sogou-articles/original/0d712459-001c-470b-9f98-0184e901deb3/0d712459-001c-470b-9f98-0184e901deb3
[e426a9f0-d83a-4b74-8f3c-708889b41472]: https://file.ai100.com.cn/files/sogou-articles/original/e426a9f0-d83a-4b74-8f3c-708889b41472/e426a9f0-d83a-4b74-8f3c-708889b41472
[f85cf8ef-cd33-4971-b8d3-c9edd4267423]: https://file.ai100.com.cn/files/sogou-articles/original/f85cf8ef-cd33-4971-b8d3-c9edd4267423/f85cf8ef-cd33-4971-b8d3-c9edd4267423
[18fa852d-5123-4568-9955-32e1b4ad9d37]: https://file.ai100.com.cn/files/sogou-articles/original/18fa852d-5123-4568-9955-32e1b4ad9d37/18fa852d-5123-4568-9955-32e1b4ad9d37
[d090936f-45f7-405f-b183-f77d61cfd9aa]: https://file.ai100.com.cn/files/sogou-articles/original/d090936f-45f7-405f-b183-f77d61cfd9aa/d090936f-45f7-405f-b183-f77d61cfd9aa
[6eecce13-29a1-47f7-b617-93bc66c80038]: https://file.ai100.com.cn/files/sogou-articles/original/6eecce13-29a1-47f7-b617-93bc66c80038/6eecce13-29a1-47f7-b617-93bc66c80038
[9d8755b8-e070-4a41-b536-e1314e76d153]: https://file.ai100.com.cn/files/sogou-articles/original/9d8755b8-e070-4a41-b536-e1314e76d153/9d8755b8-e070-4a41-b536-e1314e76d153
[cc861c84-e464-4a2f-8e38-90e3114f0ce1]: https://file.ai100.com.cn/files/sogou-articles/original/cc861c84-e464-4a2f-8e38-90e3114f0ce1/cc861c84-e464-4a2f-8e38-90e3114f0ce1
[637196d4-db08-4adf-887a-d0d5b2d5e80f]: https://file.ai100.com.cn/files/sogou-articles/original/637196d4-db08-4adf-887a-d0d5b2d5e80f/637196d4-db08-4adf-887a-d0d5b2d5e80f
[f0d393e6-3582-4444-a6bb-ea99e21703ad]: https://file.ai100.com.cn/files/sogou-articles/original/f0d393e6-3582-4444-a6bb-ea99e21703ad/f0d393e6-3582-4444-a6bb-ea99e21703ad
[05e27d23-d06a-4bf7-b9d1-f29bd55d6e75]: https://file.ai100.com.cn/files/sogou-articles/original/05e27d23-d06a-4bf7-b9d1-f29bd55d6e75/05e27d23-d06a-4bf7-b9d1-f29bd55d6e75
[AI Challenger 2018 _ PPT _--- 1]: http://mp.weixin.qq.com/s?__biz=MjM5ODkzMzMwMQ==&mid=2650408922&idx=1&sn=b15de65f00006b094a6eaaff99330bdf&chksm=becd818089ba08963ea1b2767580fc497ac7d56e02c0f92f1c29654d4fccd287d51f51a9a7a4&scene=21#wechat_redirect