Use model built using ml4ir inside of SOLR

[JnaneshPrabhu]

Hi, I am new to solr space and bumped across this library. I have been trying to use NER of search query during query_time. So for instance if I type Puma Black Shoes , the input for matching should be something of Puma <Brand> Black shoes . I have scouted the net but couldn't find any good resource available on this, I would basically want an externally trained model to be integrated inside of SOLR, so that once a model is trained, its json file is utilised inside of solr. Any ideas? Also how can i do NER using ml4ir? Can't seem to be a guide on that in documentation provided. Thanks

[jakemannix]

Hi there @JnaneshPrabhu, ml4ir in theory could be used inside of Solr, but note that it's not just a "json file" as the model: it uses Tensorflow, so it does some serious machine learning inference at query time in this case. We don't currently have "out of the box" NER, although certain kinds of NER could be handled by the text classification code in here - if you have a large quantity of training data for your corpus.

We don't currently have any "pre-trained" models. At present, this library requires the users to bring their own training data to train their models.