Podcasts are a fantastic resource of interesting and insightful opinions on a wide-range of topics. Frustratingly, it's very difficult to know what's in each episode which also makes discovery difficult. Secondly, while audio in it's original form, it's also restricting the audience to those without hearing difficulties.
What it does
Poddle takes podcast RSS submissions, checks for new episodes and then passes them through IBM's Speech to Text API. With the results of this we can provide a transcription service.
Further more, Poddle passes this transcription into the IBM Watson concept insights APIs so that we can provide a better podcast search service that's based around concepts rather than just search terms.
How we built it
We used the laravel platform as a base for the site. For the API integration, we decoupled this into an independent PHP library.
Challenges we ran into
Transcription is a slow service and we didn't get it ready in time to pass many episodes through it which means the transcription and concept search isn't visible on the demo site yet.
The IBM APIs, while reasonably documented, still took a while to get used to and understand exactly how to interact with properly.
Accomplishments that I'm proud of
Working prototype that with more time has the potential to be a true iTunes competitor for podcasts.
What's next for Poddle
Next steps for poddle are to give it time to transcribe and index more Podcasts. Also have our own spider searching for new podcasts would be a great way of increasing the library.
We also like the idea of being able to provide an alternative RSS feed of text transcriptions in article form rather than podcast.
For podcasts, it's also a challenge to know where in the podcast the topic is discussed which interests you most. With text transcription service and the timings that come back, we image being able to interact and listen to the podcast episode on Poddle - navigating directly to the sentence you wish to listen to.
Finally, as speech to text, language translation and text to speech services get more advanced we could make Podcasts even more accessible by converting them to other languages and back to audio form.