-
Notifications
You must be signed in to change notification settings - Fork 0
Hanpx20/Anchor_Clusters
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
我从每个组中选取了200个Query的样例,通过人工观察判断它们所属的领域。 1, 5, 7我看的不太确定,因此把完成的Query都放上来了。 0:医疗,健康 1*:政治,生物学 2:电子产品 3:食物,菜谱 4:人物 5*:历史 6:产品广告 7*:旅游 8:新闻 9:公司,组织 10:高校 11:人物 在目前的几次训练中,Group Weight的分布大致如下: 0 1 2 3 4 5 6 7 8 9 10 11 0.0936 0.1970 0.0949 0.0674 0.0498 0.0668 0.0440 0.0368 0.1897 0.0393 0.0680 0.0527 0.0863 0.1515 0.0952 0.0674 0.0656 0.0742 0.0543 0.0506 0.1581 0.0525 0.0805 0.0638 0.0413 0.4236 0.0496 0.0240 0.0155 0.0237 0.0095 0.0044 0.3715 0.0036 0.0178 0.0154 (这一组没加scheduler) 从小到大的顺序: 7, 9, 6, 4, 11, 5, 3, 10, 0, 2, 8, 1 7, 9, 6, 11, 4, 3, 5, 10, 0, 2, 1, 8 9, 7, 6, 11, 4, 10, 5, 3, 0, 2, 8, 1 Tier 0:1,8 Tier 1:0,2 Tier 2:3,5,10 Tier 3:4, 11 Tier 4:6,7,9 I think an explanation is that topics with high complexity and no or little proper nouns are with higher weight. 目前的训练将不同组截断到了同样的长度,如果不截断,不同组的样本数量分别为(单位:万): 0 1 2 3 4 5 6 7 8 9 10 11 34.2 49.2 27.9 30.2 18.1 50.2 43.4 56.1 32.8 39.2 27.8 33.4
About
No description, website, or topics provided.
Resources
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published