Notes for Demo page and Jupyter Notebook

The Web Page:

• The demo web page you came to this page from demonstrates a) the simple use of Word2Vec using a single seed word, and a more synthetic approach where a second seed word is added that influences the results of the first term by selecting result terms that have the highest average value with respect to the two seed terms, e.g. the average of the closest neighbour weight with seed1 and the weight with seed2.
• Once you select one of the options you get result tables. Different kinds of words demonstrate different patters of relatedness between seed terms and the result terms from word2vec.
• There is a button at the bottom of each table labelled 'Make Table'. Clicking does just that, including the ticked selected terms in the list and making a mxm table the the list against itself.
• Once you have made a table, this shows the number of SREs (paragraphs for the most part) that contain the row term and column term. Note this is not a proper cross-tabulation, as a given SRE can appear in many cells.
• Clicking the cell will start to display the SREs that make up the cell value, five at a time.
• You can click on Prev or Next to see the next or previous SRE for the selected SRE.
• You can tick the box to left of SRE to 'remember' this entry for later retrieval.
• Clicking on 'Related' will apply find other SREs that are closely related to the selected SRE.

Jupyter Notebook

• This notebook gives a more hands on approach to some of the services the demo page above is based on. You will need to install Jupyter Notebook or Jupyter Lab to proceed. This is fairly easy. Google/Bing/Whatever for instructions for doing this.
• Run Jupyter Lab, or double-click on the Demo.
• Unless you run data science sortware for python, you will have to load most of the packages in the first/second cell of the notebook. When you run the shell, the notebook will have a listing of what the error was.