# Search functionality for labels

**URL:** <https://support.prodi.gy/t/search-functionality-for-labels/4639>\
**Category:** Uncategorized\
**Tags:** usage, front-end\
**Created:** [August 31, 2021, 2:21pm UTC](https://support.prodi.gy/t/search-functionality-for-labels/4639 "2021-08-31T14:21:42Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![AnnaGuledMade](https://sea2.discourse-cdn.com/flex020/user_avatar/support.prodi.gy/annaguledmade/32/2378_2.png) [@AnnaGuledMade](https://support.prodi.gy/u/AnnaGuledMade)\
**Post date:** [August 31, 2021, 2:21pm UTC](https://support.prodi.gy/t/search-functionality-for-labels/4639/1 "2021-08-31T14:21:42Z")

</div>

Hi there,  
we have a very large number of labels that we'll use for annotations.  
However, I cannot find a search functionality in the documentation.

Can you please advise?  
Many thanks

---

<div class="post-metadata">

**Author:** ![ines](https://sea2.discourse-cdn.com/flex020/user_avatar/support.prodi.gy/ines/32/3_2.png) [@ines](https://support.prodi.gy/u/ines)\
**Post date:** [August 31, 2021, 3:24pm UTC](https://support.prodi.gy/t/search-functionality-for-labels/4639/2 "2021-08-31T15:24:20Z")

</div>

Hi! I think I may have already answered you via email (assuming you/your colleague emailed us earlier about span categorization and search functionality for thousands of labels). Copying over my reply, also in case it's useful for others 🙂

You can change the label view to a dropdown instead of a list of labels (using the `label_style` config setting) – however, if your goal is to train a span categorizer and you ended up with this many labels, we'd usually recommend rethinking your label scheme and breaking down the task differently so you're able to annotate and train more efficiently.

With this many labels, the annotators will otherwise have to think about 6k options on every example and search for the correct one, which can get pretty slow and lead to mistakes and inconsistent data. In order to predict a large number of labels, you also need sufficient examples of every label in context (which are naturally unevenly distributed). So you're potentially making your life a lot harder than it should be.

If your labels are hierarchical, one common solution is to start by annotating the top-level categories (which is usually the most important distinction) and then have separate processes for selecting the more fine-grained labels. This also lets you train and iterate a lot quicker. Also see this thread for some background (it's about NER but the considerations for SpanCat are pretty similar):

> [@NER with dozens of entities](https://support.prodi.gy/t/ner-with-dozens-of-entities/4139/3):
>
> As a more general comment/question: Is your goal to actually train an NER model? If so, are you sure you want to structure your annotation task and model this way and have that many labels? You might be making your life a lot harder this way, because you'll need significantly more data and make sure you have enough examples per label so the model can distinguish between them. If the labels are hierarchical, you'd also be encoding multiple aspects within each annotation that the model has to lear…