creating a new derive concept have problem

[greentownwinwinwin]

If I try to follow each step, when I'm creating a new derive concept, an error has occured and need to refresh the session. Refresh the session not resolve the problem, so often I lost everything created and need to restart again

[Chenglong-MS]

Ooops, sorry about that --- the derived concept hasn't been well tested since it was a design from the earlier version. I'll take a look and fix issues to prevent erroneous states.

[greentownwinwinwin]

earlier version

Ooops, sorry about that --- the derived concept hasn't been well tested since it was a design from the earlier version. I'll take a look and fix issues to prevent erroneous states.

The earlier version of table could select data, but not in 0.1.6.1. Could you please fix this issue?

[greentownwinwinwin]

earlier version

Ooops, sorry about that --- the derived concept hasn't been well tested since it was a design from the earlier version. I'll take a look and fix issues to prevent erroneous states.

The earlier version of table could select data, but not in 0.1.6.1. Could you please fix this issue?

The paper states that after the formulator is executed, multiple tables will be generated for selection. However, in actual operation, only one visualized table is generated. I wonder if this is a problem with my operation or if it is a limitation that only one table can be generated.

[Chenglong-MS]

I believe you were referring to the first Data Formulator paper, in our new design Data Formulator 2, we changed the design to include only one visualizations based on user intent -- this design bases on our observations from the user study that multiple alternative can increase cognitive loads from the users to identify differences (sometimes subtle) among different results (despite there are some benefits when top-ranked results is not quite right). Generating multiple charts also slow down AI outputs. Happy to share more behind our design.

Select data was also a feature from the old Data Formulator (2023) to support programming by example, wondering why that feature would be useful in the current version? We may consider adding it back if it is useful.

[greentownwinwinwin]

I believe you were referring to the first Data Formulator paper, in our new design Data Formulator 2, we changed the design to include only one visualizations based on user intent -- this design bases on our observations from the user study that multiple alternative can increase cognitive loads from the users to identify differences (sometimes subtle) among different results (despite there are some benefits when top-ranked results is not quite right). Generating multiple charts also slow down AI outputs. Happy to share more behind our design.

Select data was also a feature from the old Data Formulator (2023) to support programming by example, wondering why that feature would be useful in the current version? We may consider adding it back if it is useful.

在table中选择数据，可以简化数据清洗的过程，而不用再给promt去生成，因为有一些数据表的colum其实是无效的，其次，生成多个可视化数据表有助于非专业数据分析师用户更好的选择，因为非专业用户需要参考比较

[Chenglong-MS]

• 选择数据你指的是直接在表中edit数据吗？
• 没错，生成多个可视化在recommendation场景有用，我们看一下怎么能在data threads里面比较自然的展示出来

[greentownwinwinwin]

• 选择数据你指的是直接在表中edit数据吗？
• 没错，生成多个可视化在recommendation场景有用，我们看一下怎么能在data threads里面比较自然的展示出来

之前的版本，table可以选择数据，但是最新版本没有，其他的表都可以选择，只有table不可以

[Chenglong-MS]

还是不太明白选择数据指的是什么... 是下图这样的select吗？

[Chenglong-MS]

Also, derived concept bug is now fixed in the dev branch. Feel free to test out, we'll merge to main next week:

https://github.com/microsoft/data-formulator/tree/dev

[greentownwinwinwin]

还是不太明白选择数据指的是什么... 是下图这样的select吗？

是的 还有就是table没有右侧的拖拽选择

[greentownwinwinwin]

我想问一下前端是用什么框架写的，前端的dataformulator.js文件特别大，我想进行一些修改，但是不知道从何下

[Chenglong-MS]

前端使用React + Redux写的，你可以参考https://github.com/microsoft/data-formulator/blob/main/DEVELOPMENT.md 来进行修改。

还有就是你说的table不能选择是因为那是derived data生成的用来解释derived concept用的example data，并不是最终table。如果使用chart-builder里面的data formulation，生成出来的table会显示在main panel，这是可以选择的。