Creating a wordvector model combining words from other models

I have two different word vector models created using word2vec algorithm . Now issue i am facing is few words from first model is not there in second model . I want to create a third model from two different word vectors models where i can use word vectors from both models without loosing meaning and the context of word vectors.

Can I do this, and if so, how?

Answer1:

You could potentially translate the vectors for the words only in one model to the other model's coordinate space, using other shared words to learn a translation-function.

There's a facility to do this in recent gensim versions – see the TranslationMatrix tool. There's a demo Jupyter notebook included in the docs/notebooks directory, viewable online at:

https://github.com/RaRe-Technologies/gensim/blob/develop/docs/notebooks/translation_matrix.ipynb

You'd presumably take the larger model (or whichever one is thought to be better, perhaps because it was trained on more data), and translate the smaller number of words its missing into its space. You'd use as many common-reference 'anchor' words as is practical.

人吐槽 人点赞

Recommend

  • Pass dropdown value to code behind static method
  • In C++, is it possible to have a class inherit from one of its member classes?
  • XPath evaluate a query and retrieve the value
  • .cs files can't be opened anymore in VS 2013 Professional Update 4
  • Replacing the Translator service in Symfony 3
  • Faster way to compute likelihood of sequence?
  • Why do class member functions defined outside the class (but in header file) have to be inlined?
  • The exact mechanism of mapping WhyML into SMT logic
  • When does initialisation of global variables happen?
  • Copying different structs to byte arrays
  • Moving Object in ScrollViewer
  • How do I specify custom wording in a will_paginate view helper?
  • How to use a decaying learning rate with an estimator in tensorflow?
  • Getting coordinates of a component in java
  • Connect Node.js with Oracle on Windows platform
  • Extract All Possible Paths from Expression-Tree and evaluate them to hold TRUE
  • uml Composition relationships to RDF and OWL
  • Ubuntu and bcrypt
  • Filter strings with regex before casting to numeric
  • How to detect interior vertices in groups of 2d polygons? (E.g. ZIP Codes to determine a territory)
  • How do I configure context broker accept post requests from my remote sensor?
  • D3 get axis values on zoom event
  • x64 applications using gdi+: what are the consequences on performance?
  • How to define custom class, title, and target in Link Browser for content elements and the new rte_c
  • Admob requires api-13 or later can I not deploy on old API-8 phones?
  • jQuery .attr() and value
  • Javascript Callbacks with Object constructor
  • 0x202A in filename: Why?
  • Angular 2 constructor injection vs direct access
  • How does Linux kernel interrupt the application?
  • Comment

    用户名: 密码:
    验证码: 匿名发表

    你可以使用这些语言

    查看评论:Creating a wordvector model combining words from other models