Update keyedvectors.py #3311
Update keyedvectors.py #3311
Conversation
Following the discussion in the Gensim google group at https://groups.google.com/u/1/g/gensim/c/gCJK7-2QwRw, here is my suggestion for modifying the first part of the explanation of the function evaluate_word_pairs(). Thank you!
piskvorky
added
feature
|
Hi, do I need to make any changes for it to get approved? Thank you! |
gensim/models/keyedvectors.py
Outdated
| """Compute correlation of the model with human similarity judgments. | ||
| """Compute correlation of the model with human annotated or taxonomy‐based semantic similarity measures. | ||
| More information on the evaluation of word embeddings through gold benchmark similarity data | ||
| can be found at * https://link.springer.com/article/10.1007/s12559-021-09987-7. |
gensim/models/keyedvectors.py
Outdated
| See `test/test_data/wordsim353.tsv` as example of human similarity judgments, | ||
| and test/test_data/HSS4570.tsv as an example of taxonomy‐based semantic similarity. | ||
| What you should use and why: | ||
| - human similarity judgments: |
There was a problem hiding this comment.
Is this a nested list, with several nested :?
gensim/models/keyedvectors.py
Outdated
| What you should use and why: | ||
| - human similarity judgments: | ||
| `why`: they are the most used benchmarks in word embedding evaluation. | ||
| `why not`: |
There was a problem hiding this comment.
Why literal formatting for this list heading?
You are right, there are too many nested lists for it to be clear. I modified it to leave the parameters section clean and to add after the returns section another section explaining when to use the human similarity judgments and the taxonomy‐based semantic similarity measures with a nested list of why and why not for each. I will attach a screenshot of the rendered HTML for the section "What you should use and why" that should be added after the "Returns" section:
|
Following the discussion in the Gensim google group at https://groups.google.com/u/1/g/gensim/c/gCJK7-2QwRw, here is my suggestion for modifying the first part of the explanation of the function evaluate_word_pairs(). Thank you!