# Tools helping creating good datasets

**URL:** <https://forum.rasa.com/t/tools-helping-creating-good-datasets/1457>\
**Category:** Rasa Open Source\
**Created:** [September 27, 2018, 2:22pm UTC](https://forum.rasa.com/t/tools-helping-creating-good-datasets/1457 "2018-09-27T14:22:58Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![datistiquo](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/datistiquo/32/124_2.png) [@datistiquo](https://forum.rasa.com/u/datistiquo)\
**Post date:** [September 27, 2018, 2:22pm UTC](https://forum.rasa.com/t/tools-helping-creating-good-datasets/1457/1 "2018-09-27T14:22:58Z")

</div>

How do you create in practice well balanced data sets for intent and entity classification? Do you offer some tools for that in Rasa Platform @akelad? Because a tool for just adding data is really not enough. You have to take care of balancing different sentence types for each entity and intent classification. I have several thousand of example and struggle to find the proper ones and deleting similiar sentences. Would it be a good strategy for practice to have a tool where you have a look over your structures for sentences? I imagine a simple bag of word topic clustering to cluster my examples and choose the right ones for the tensorlfow embedding. For entities I think a simple statistics over context words would be enough. So you can see how many example you have of each type of context for entities and you can avoiding overfitting. I think that would be a good approach?