Before introducing this code, I want to insert a couple caveats:
- Forvo is an excellent resource for language learners and I wish the company the best of luck in commercializing word and phrase pronunciations. This code is not meant to deprive them of resources to further their misson.
- They have an API that in principle should allow developers to access pronunciations. In fact, I'm a paid monthly subscriber to that API. But it is often very slow and frequently is down completely without any status information or response from the support staff.
- This is meant as an example for you to adapt. It only works for Russian pronunciations right now.
There are other projects - mostly Anki add-ons - that scrape pronunciations in this same fashion. The intent here is to do the same thing without integration into Anki. This was also a platform for me to learn to integrate Selenium with Beautiful Soup 4. For reasons unclear to me, despite using a browser user-agent string, some of my requests from Python to a Forvo pronunciation page resulted in errors. Therefore, I brought in Selenium to completely mimic the browser experience when viewed from the server side. This is all a little slower than using the API, as long as the API is functioning as it did in the past. Now it's often extremely slow or down.
One unique feature of this Forvo scraper is that it seeks to prioritize pronunciation through a simple scoring system. You can change FAVS to whatever users are your favourites. Then the Pronunciation class attempts to rank the pronunciations by favourite user and number of votes.
To use, you'll need to install Beautiful Soup 4, Selenium and anything else you don't have. Then download the code in forvo_scraper_example.py and make it executable. Then it's just forvo_scraper_example.py --word 'кошка' --dest '/your/destination/directory'