Wikipedia2Vec: Optimized Implementation for Learning Embeddings from Wikipedia

Re:Wikipedia2Vec: Optimized Implementation for Learning Embe... by crucialfelix on Sunday December 23, 2018 @02:42PM (#18747903)

Perfect timing for me. I've been working on a project that needs multilingual vectorizing and entity recognition. I've been using dbpedia and yago queries.

Given an input text, what would be a good way to extract a list of entities? The word sequence should be usable to determine which is an entity or just a word.

Is there the possibility to do fine tuning a la Bert or Elmo?

[ Parent | on HN ]

Re:Wikipedia2Vec: Optimized Implementation for Learning Embe... by ikuyamada on Sunday December 23, 2018 @08:20PM (#18749610)

My past paper describes an entity linking method based on Wikipedia2Vec: https://arxiv.org/abs/1601.01343

You need to extract entity names using an NER software (e.g., SpaCy, Stanford NER), and resolve the names to knowledge base entities using the entity linking method.

[ Parent | on HN ]

Re:Wikipedia2Vec: Optimized Implementation for Learning Embe... by currymj on Sunday December 23, 2018 @06:54PM (#18749251)

You can use the TAGME API to do that, it’s state of the art as far as I know. If you want to implement it yourself I think it’s pretty painful, though.

No reason you can’t fine tune these on your task, that’s true of any word embeddings not just Bert or Elmo.

[ Parent | on HN ]