# Multilingual NER pretrained model fine tuning

**URL:** </static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fmultilingual-ner-pretrained-model-fine-tuning%2F65186%3E%5C
**Category:** Models\
**Created:** [December 9, 2023, 8:29pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fmultilingual-ner-pretrained-model-fine-tuning%2F65186 "2023-12-09T20:29:25Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![Mohamedabokahf](https://avatars.discourse-cdn.com/v4/letter/m/6de8d8/32.png) [@Mohamedabokahf](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FMohamedabokahf)%5C
**Post date:** [December 9, 2023, 8:29pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fmultilingual-ner-pretrained-model-fine-tuning%2F65186%2F1 "2023-12-09T20:29:25Z")

</div>

**Hallo**  
I want to fine tun a pretrained model for named entity recognition. I have chosen Roberta base to fine tun, I had the problem that the dataset contains some Russian words, which the model is not able to understand while the Tokenisation. I thought about adding these words (it depends on how many word I have) to the tokenizer, but I am not sure if the model will understand that those are ORG, OTH etc. so what do you recommend as a solution for this case. The dataset is .tsv file includes German words with numbers and labels I have made them in pandas data frame index, token, label then I get the sentence based on the boundars for example if index 1 this is a start of sentence and if index 1 this is end of the last sentence and beginning of new one. This is one of the sentences : (NEWSru.ua / : Политисполком СПУ отказал Морозу в отставке Die SPU legte wie auch vier weitere Parteien beim Obersten Gericht ’   
'der Ukraine Beschwerde gegen den Ablauf der Wahl ein und behauptet es sei bei der Auszählung zu Unregelmäßigkeiten gekommen .) all the Russian words are supposed to be I-OTH in the truth label.