gelectra-large for Extractive QA

Overview

Language model: gelectra-large-germanquad
Language: German
Training data: GermanQuAD train set (~ 12MB)
Eval data: GermanQuAD test set (~ 5MB)
Code: See an example extractive QA pipeline built with Haystack
Infrastructure: 1x V100 GPU
Published: Apr 21st, 2021

Details

We trained a German question answering model with a gelectra-large model as its basis.
The dataset is GermanQuAD, a new, German language dataset, which we hand-annotated and published online.
The training dataset is one-way annotated and contains 11518 questions and 11518 answers, while the test dataset is three-way annotated so that there are 2204 questions and with 2204·3−76 = 6536 answers, because we removed 76 wrong answers.

See https://deepset.ai/germanquad for more details and dataset download in SQuAD format.

Hyperparameters

batch_size = 24
n_epochs = 2
max_seq_len = 384
learning_rate = 3e-5
lr_schedule = LinearWarmup
embeds_dropout_prob = 0.1

Usage

In Haystack

Haystack is an AI orchestration framework to build customizable, production-ready LLM applications. You can use this model in Haystack to do extractive question answering on documents. To load and run the model with Haystack:

# After running pip install haystack-ai "transformers[torch,sentencepiece]"

from haystack import Document
from haystack.components.readers import ExtractiveReader

docs = [
    Document(content="Python is a popular programming language"),
    Document(content="python ist eine beliebte Programmiersprache"),
]

reader = ExtractiveReader(model="deepset/gelectra-large-germanquad")
reader.warm_up()

question = "What is a popular programming language?"
result = reader.run(query=question, documents=docs)
# {'answers': [ExtractedAnswer(query='What is a popular programming language?', score=0.5740374326705933, data='python', document=Document(id=..., content: '...'), context=None, document_offset=ExtractedAnswer.Span(start=0, end=6),...)]}

For a complete example with an extractive question answering pipeline that scales over many documents, check out the corresponding Haystack tutorial.

In Transformers

from transformers import AutoModelForQuestionAnswering, AutoTokenizer, pipeline

model_name = "deepset/gelectra-large-germanquad"

# a) Get predictions
nlp = pipeline('question-answering', model=model_name, tokenizer=model_name)
QA_input = {
    'question': 'Why is model conversion important?',
    'context': 'The option to convert models between FARM and transformers gives freedom to the user and let people easily switch between frameworks.'
}
res = nlp(QA_input)

# b) Load model & tokenizer
model = AutoModelForQuestionAnswering.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)

Performance

We evaluated the extractive question answering performance on our GermanQuAD test set. Model types and training data are included in the model name. For finetuning XLM-Roberta, we use the English SQuAD v2.0 dataset. The GELECTRA models are warm started on the German translation of SQuAD v1.1 and finetuned on GermanQuAD. The human baseline was computed for the 3-way test set by taking one answer as prediction and the other two as ground truth.