# Rasa Nlu custom pipeline

**URL:** <https://forum.rasa.com/t/rasa-nlu-custom-pipeline/4062>\
**Category:** Rasa Open Source\
**Created:** [January 5, 2019, 2:44pm UTC](https://forum.rasa.com/t/rasa-nlu-custom-pipeline/4062 "2019-01-05T14:44:58Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![vigneshgig](https://avatars.discourse-cdn.com/v4/letter/v/eada6e/32.png) [@vigneshgig](https://forum.rasa.com/u/vigneshgig)\
**Post date:** [January 5, 2019, 2:44pm UTC](https://forum.rasa.com/t/rasa-nlu-custom-pipeline/4062/1 "2019-01-05T14:44:58Z")

</div>

Hi , I want to integrate the rasa nlu with custom NER like BERT, Can anyone suggest how to do it.there are no tutorial regarding custom pipeline, So far i seen little bit information in rasa nlu documentation.So please help me out.

---

<div class="post-metadata">

**Author:** ![nahidalam](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/nahidalam/32/91_2.png) [@nahidalam](https://forum.rasa.com/u/nahidalam)\
**Post date:** [January 11, 2019, 3:23pm UTC](https://forum.rasa.com/t/rasa-nlu-custom-pipeline/4062/2 "2019-01-11T15:23:46Z")

</div>

This is an interesting idea. You have to create the BERT model (or use anything existing) and make it part of the pipeline.

Does this link help? [https://rasa.com/docs/nlu/0.13.8/customcomponents/](https://rasa.com/docs/nlu/0.13.8/customcomponents/)

An interesting blog of someone adding a new model with Rasa is here [https://scalableminds.com/blog/MachineLearning/2018/08/rasa-universal-sentence-encoder/](https://scalableminds.com/blog/MachineLearning/2018/08/rasa-universal-sentence-encoder/)

Hope this is helpful.

---

<div class="post-metadata">

**Author:** ![vigneshgig](https://avatars.discourse-cdn.com/v4/letter/v/eada6e/32.png) [@vigneshgig](https://forum.rasa.com/u/vigneshgig)\
**Post date:** [January 12, 2019, 5:56am UTC](https://forum.rasa.com/t/rasa-nlu-custom-pipeline/4062/3 "2019-01-12T05:56:57Z")

</div>

Thanks, @nahidalam one more question in rasa is there any auto greeting feature.

---

<div class="post-metadata">

**Author:** ![nahidalam](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/nahidalam/32/91_2.png) [@nahidalam](https://forum.rasa.com/u/nahidalam)\
**Post date:** [January 12, 2019, 7:23pm UTC](https://forum.rasa.com/t/rasa-nlu-custom-pipeline/4062/4 "2019-01-12T19:23:22Z")

</div>

Not that I know off…

---

<div class="post-metadata">

**Author:** ![FTAsr](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/ftasr/32/5290_2.png) [@FTAsr](https://forum.rasa.com/u/FTAsr)\
**Post date:** [October 4, 2019, 7:33pm UTC](https://forum.rasa.com/t/rasa-nlu-custom-pipeline/4062/5 "2019-10-04T19:33:19Z")

</div>

None of the links you provided work or contain any relevant information on custom pipelines. Is Rasa python coding out of the window now? How can one simply create a small pipeline, train it on data from within python code (no shell) and test the model on a single input?

I’m trying the following code snippet from the old versions and it does not work… none of the libraries exist anymore:

```
from rasa_nlu.training_data import load_data
from rasa_nlu.config import RasaNLUModelConfig
from rasa_nlu.model import Trainer
from rasa_nlu import config
from rasa_nlu.model import Metadata, Interpreter

training_data = load_data('data/examples/rasa/demo-rasa.json')
trainer = Trainer(config.load("sample_configs/config_spacy.yml"))
trainer.train(training_data)
model_directory = trainer.persist('./projects/default/') # Returns the directory the model is stored in

# where `model_directory points to the folder the model is persisted in
interpreter = Interpreter.load(model_directory)

print(interpreter.parse(u"The text I want to understand"))
```