当前位置：网站首页>Huawei cloud modelarts text classification - takeout comments

Huawei cloud modelarts text classification - takeout comments

2022-07-05 21:43:00 【InfoQ】

1、 Data brief

This data set is the takeout review data set , Contains about 4000 A positive comment 、8000 Negative comments . Data set text objects and labels are stored in the following different files ：

comment.txt

comment_result.txt

2、 Applicable algorithms

This data set can be used for the following AI Gallery Chinese text classification algorithm for training Chinese text classification -Bert：

AI Gallery_ Algorithm _ Model _ Cloud market - Hua Wei Yun

be based on BERT Text classification algorithm of pre training model , Support single label text classification . Pre training model based on

Google released BERT BASE Model

. Users need to complete the annotation on the data management platform , The algorithm will load the pre training model and do migration learning on the user data set . The model generated after training can be directly in ModelArts The platform is deployed as an online service , It also supports the use of CPU、GPU、Ascend310 Reasoning .Ascend310 Reasoning requires configuration export_d_model Parameter is True.

3、 Usage method

a、 Download this dataset to your OBS In the barrel

b、 stay ModelArts“ Data management - Data sets ” Create a text classification dataset （ Be careful ： Data set input location 、 Data set output location selection 2 An empty one OBS route ）

c、 In the upper right corner of the dataset details page “ Import ”, Import mode selection “ Object storage service （OBS） Catalog ”, Select the above download path for the directory location , Mode selection “ Text and annotation are separated ”

d、 After the import is completed , Will be displayed 11987 Data marked , Click on the top right corner of the page “ Release ”, Data according to 0.8:0.2 Proportional segmentation

e、 Subscribe to text classification algorithm , Select the text classification data set published above for training