**Input:** Supports over 100 Languages. See [List of supported languages](https://github.com/google-research/bert/blob/master/multilingual.md#list-of-languages) for all available.
**Purpose:** This module takes a search query [1] and a passage [2] and calculates if the passage matches the query.
It can be used as an improvement for Elasticsearch Results and boosts the relevancy by up to 100%.
**Architecture:** On top of BERT there is a Densly Connected NN which takes the 768 Dimensional [CLS] Token as input and provides the output ([Arxiv](https://arxiv.org/abs/1901.04085)).
**Output:** Just a single value between between 0-1
## Intended uses & limitations
Both query[1] and passage[2] have to fit in 512 Tokens.
As you normally want to rerank the first dozens of search results keep in mind the inference time.
This Model can be used as a drop-in replacement in the [Nboost Library](https://github.com/koursaros-ai/nboost)
Through this you can directly improve your Elasticsearch Results without any coding.
## Training data
This model is trained using the [**Microsoft MS Marco Dataset**](https://microsoft.github.io/msmarco/"Microsoft MS Marco"). This training dataset contains approximately 400M tuples of a query, relevant and non-relevant passages. All datasets used for training and evaluating are listed in this [table](https://github.com/microsoft/MSMARCO-Passage-Ranking#data-information-and-formating). The used dataset for training is called *Train Triples Large*, while the evaluation was made on *Top 1000 Dev*. There are 6,900 queries in total in the development dataset, where each query is mapped to top 1,000 passage retrieved using BM25 from MS MARCO corpus.
## Training procedure
The training is performed the same way as stated in this [README](https://github.com/nyu-dl/dl4marco-bert"NYU Github"). See their excellent Paper on [Arxiv](https://arxiv.org/abs/1901.04085).
We changed the BERT Model from an English only to the default BERT Multilingual uncased Model from [Google](https://huggingface.co/bert-base-multilingual-uncased).
Training was done 400 000 Steps. This equaled 12 hours an a TPU V3-8.
## Eval results
We see nearly similar performance than the English only Model in the English [Bing Queries Dataset](http://www.msmarco.org/). Although the training data is English only internal Tests on private data showed a far higher accurancy in German than all other available models.
Fine-tuned Models | Dependency | Eval Set | Search Boost<ahref='#benchmarks'> | Speed on GPU
