# NER at the Inference Time

**URL:** <https://discuss.huggingface.co/t/ner-at-the-inference-time/15855>\
**Category:** 🤗Transformers\
**Created:** [March 18, 2022, 4:35pm UTC](https://discuss.huggingface.co/t/ner-at-the-inference-time/15855 "2022-03-18T16:35:27Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![OnclusiveML](https://avatars.discourse-cdn.com/v4/letter/o/ea666f/32.png) [@OnclusiveML](https://discuss.huggingface.co/u/OnclusiveML)\
**Post date:** [March 18, 2022, 4:35pm UTC](https://discuss.huggingface.co/t/ner-at-the-inference-time/15855/1 "2022-03-18T16:35:27Z")

</div>

Apology in advance if this question might have already been asked. However, I have not been able to find a convincing answer or the optimum way to deal with this issue.

To my understanding, NER makes a prediction at a token level. Since BERT is using a sub-word tokenizer, it is entirely possible that some part of the word won’t get labeled or we have a different label within the same word. Both of these are undesirables because in the end we want the final result to be NER on a word, not token, level.

For example, see [link](https://huggingface.co/dslim/bert-base-NER?text=Seven+of+the+men+are+from+so-called+%22red-flagged%22+countries%2C+including+Egypt%2C+Turkey%2C+Georgia%2C+Pakistan+and+Mali.Her+eighth+husband%2C+Rashid+Rajput%2C+was+deported+in+2006+to+his+native+Pakistan+after+an+investigation+by+the+Joint+Terrorism+Task+Force.If+convicted%2C+Barrientos+faces+up+to+four+years+in+prison.++Her+next+court+appearance+is+scheduled+for+May+18.).

Barr(PER) ien(O) tos(PER)

This should have been Barr(PER) ien(PER) tos(PER).

Another more confusing prediction [here](https://huggingface.co/dslim/bert-base-NER?text=27+Colors+for+Main+Fabric+available+in+singles+%26+together+in+Fatpack+Sweater)

F(MISC) abric (O) … Fat (LOC) pack Sweater (ORG)

We have inconsistent token production within the same words.

So my questions are the following

1. How can I best convert token-level NER labels to word-level labels? What is the best policy to deal with inconsistent token-level prediction within the same word? Is there a standard way to do this? Has this already been implemented in the huggingface library.

2. Should not there be a way to force the model to recognize that those three tokens came from the same word so they need to have the same token-level label in the first place?

Any suggestions are greatly appreciated. Thank you!