NLP Summarizer Model for ES 8.2.2

sim · June 8, 2022, 9:15pm

I wonder if ES 8.2.2 already supports a PyTorch NLP model for summarization?

Eg. sshleifer/distilbart-cnn-12-6 · Hugging Face

It's not an option in eland's current supported task types.

It is a generative model, ie. given a long text, will produce a short text.

If it cannot be formulated with current eland, could I do it programmatically like this way?

I'd be interested to know what's currently possible, and if not yet, when will it be available?

That summarizer is BART model.

Thanks,
Sim

dkyle · June 9, 2022, 8:44am

Hi Sim

Summarization isn't a task that is supported in any version of ES yet. I opened an GH issue for the enhancement requests at Support NLP summarization models · Issue #87548 · elastic/elasticsearch · GitHub, please upvote.

The good news is that Question Answering models will be supported in 8.3 and new features are being added all the time.

github.com/elastic/elasticsearch

[ML] adds new question_answering NLP task for extracting answers to questions from a document

elastic:master ← benwtrent:feature/ml-add-question-answering-nlp-task

opened 04:21PM - 18 Apr 22 UTC

benwtrent

+1999 -80

This commit adds a new `question_answering` task. The `question_answering` ta…sk allows supplying a `question` in the inference config update. When storing the model config for inference: ``` "inference_config": { "question_answering": { "tokenization": {...}, // tokenization settings, recommend doing 386 max sequence length with 128 span, and no truncating "max_answer_length": 15 // the max answer length to consider } } ``` Then when calling `_infer` or running with in a pipeline, add the `question` you want answered given the context provided by the document text ``` { "docs":[{ "text_field": <some long text field to extract answer}], "inference_config": { "question_answering": { "question": <Question desiring answer> } } } ``` The response then looks like: ``` { "predicted_value": <string subsection of the document that is the answer> "start_offset": <Char offset in document to start>, "end_offset": <char offset end of the answer, "prediction_probability": <prediction score> } ``` Some models tested: - https://huggingface.co/distilbert-base-cased-distilled-squad - https://huggingface.co/bert-large-uncased-whole-word-masking-finetuned-squad - https://huggingface.co/deepset/electra-base-squad2 - https://huggingface.co/deepset/tinyroberta-squad2

sim · June 9, 2022, 11:17pm

Nice @dkyle! Upvoted.

That's cool, having Q&A in 8.3.

Before you saying that, I thought it's already in 8.2.2 because eland has question_answering task_type option already.

system · July 7, 2022, 11:18pm

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.