[Question]NLP text classification sample?

[LangDaoAI]

Hi Team,

I want to know whether has any NLP text classification sample about the evidently?

Thanks!

[LangDaoAI]

e.g. I am using transformer related pretrained models (bert, roborta ...) to do sentimental classification based on text data. Then, how evidently can monitor these models and text data drift ?

[LangDaoAI]

That is to say: how do we measure data drift (NLP、CV) without structured features?

[SangamSwadiK]

That is to say: how do we measure data drift (NLP、CV) without structured features?

Hey! @LangDaoAI I guess you could use dimensionality reduction followed by KS test or any other test to measure the drift.

[LangDaoAI]

That is to say: how do we measure data drift (NLP、CV) without structured features?

Hey! @LangDaoAI I guess you could use dimensionality reduction followed by KS test or any other test to measure the drift.

Hi, @SangamSwadiK in fact, in my scenes, most of them uses transformered-based models for text classfication and image classfication, how do we measure data drift (NLP、CV) without structured features using evidently ?

[LangDaoAI]

I think it's a little strange, has no one asked the same question previously?

[elenasamuylova]

Hi @LangDaoAI, right now, Evidently only natively supports tabular data as inputs.

Supporting text data is in our mid-term roadmap. This is a feature request that comes up regularly, so we will definitely address it. However, we want to build up some of the core features for tabular data first and would need to perform some additional research before we implement drift detection, etc. for unstructured data. Will get there!

[LangDaoAI]

Thanks detailed reply! I see. Get Outlook for Android<https://aka.ms/AAb9ysg>

…

________________________________ From: elenasamuylova ***@***.***> Sent: Wednesday, October 19, 2022 3:08:37 AM To: evidentlyai/evidently ***@***.***> Cc: LangDaoAI ***@***.***>; Mention ***@***.***> Subject: Re: [evidentlyai/evidently] [Question]NLP text classification sample? (Issue #392) Hi @LangDaoAI<https://github.com/LangDaoAI>, right now, Evidently only natively supports tabular data as inputs. Supporting text data is in our mid-term roadmap. This is a feature request that comes up regularly, so we will definitely address it. However, we want to build up some of the core features for tabular data first and would need to perform some additional research before we implement drift detection, etc. for unstructured data. Will get there! — Reply to this email directly, view it on GitHub<#392 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AT7ZPWOO6T4LR2RD75BIA5LWD3YTLANCNFSM6AAAAAARHSRRTM>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[elenasamuylova]

Hi @LangDaoAI, as a quick note: we've recently released raw text data support in Evidently. You can read more here: https://www.evidentlyai.com/blog/evidently-data-quality-monitoring-and-drift-detection-for-text-data

[LangDaoAI]

Grest News! 获取Outlook for Android<https://aka.ms/AAb9ysg>

…

________________________________ From: elenasamuylova ***@***.***> Sent: Tuesday, January 31, 2023 1:14:40 AM To: evidentlyai/evidently ***@***.***> Cc: LangDaoAI ***@***.***>; Mention ***@***.***> Subject: Re: [evidentlyai/evidently] [Question]NLP text classification sample? (Issue #392) Hi @LangDaoAI<https://github.com/LangDaoAI>, as a quick note: we've recently released raw text data support in Evidently. You can read more here: https://www.evidentlyai.com/blog/evidently-data-quality-monitoring-and-drift-detection-for-text-data ― Reply to this email directly, view it on GitHub<#392 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AT7ZPWI64KGDCCWAAK4RNILWU7ZIBANCNFSM6AAAAAARHSRRTM>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[LangDaoAI]

Hi @LangDaoAI, as a quick note: we've recently released raw text data support in Evidently. You can read more here: https://www.evidentlyai.com/blog/evidently-data-quality-monitoring-and-drift-detection-for-text-data

Hi @elenasamuylova , noting the following:
"What’s more, you can pass multi-modal data that combines features of different types in a single dataset."
CV data also supported?

Thanks!

[LangDaoAI]

This new feature and design/implementation is really awesome! However, after I have read the blog , I felt a litter confused for "Drift" true meaning. Although we can carefully read https://arxiv.org/pdf/1810.11953.pdf to spend some time or effort on understand the point, I always want some more clean explanations about "Drift" and even happened on real world.

Thanks!