:mod:`capreolus.reranker.POSITDRMM`
===================================

.. py:module:: capreolus.reranker.POSITDRMM


Module Contents
---------------

Classes
~~~~~~~

.. autoapisummary::

   capreolus.reranker.POSITDRMM.POSITDRMM_basic
   capreolus.reranker.POSITDRMM.POSITDRMM_class
   capreolus.reranker.POSITDRMM.POSITDRMM



.. data:: logger
   

   

.. data:: device
   

   

.. py:class:: POSITDRMM_basic(extractor, pipeline_config)

   Bases: :class:`torch.nn.Module`

   .. method:: init_hidden(self)


   .. method:: forward(self, sentence, query_sentence, query_idf, extra)



.. py:class:: POSITDRMM_class(extractor, config)

   Bases: :class:`torch.nn.Module`

   .. method:: forward(self, query_sentence, query_idf, pos_sentence, neg_sentence, posdoc_extra, negdoc_extra)


   .. method:: test_forward(self, query_sentence, query_idf, pos_sentence, extras)



.. data:: dtype
   

   

.. py:class:: POSITDRMM

   Bases: :class:`capreolus.reranker.Reranker`

   .. attribute:: description
      :annotation: = Ryan McDonald, George Brokos, and Ion Androutsopoulos. 2018.
                        Deep Relevance Ranking Using Enhanced Document-Query Interactions. In EMNLP'18.

      

   .. attribute:: module_name
      :annotation: = POSITDRMM

      

   .. method:: build_model(self)


   .. method:: score(self, data)


   .. method:: test(self, data)


   .. method:: zero_grad(self, *args, **kwargs)


   .. method:: get_searcher_scores(self, qids, doc_ids)


   .. method:: clean(self, text)

      Remove pad tokens from the text


   .. method:: get_bigrams(self, text)


   .. method:: get_bigram_match_count(self, query, doc)


   .. method:: get_exact_match_count(self, query, single_doc, query_idf)


   .. method:: get_exact_match_stats(self, query_idf_batch, query_batch, doc_batch)

      The extra 4 features that must be combined to the relevant score.
      See https://www.aclweb.org/anthology/D18-1211.pdf section 4.1 beginning.