Podcast Search Service
December 18, 2015
I read “Podcasting’s Search Problem Could be Solved by This Spanish Startup.” According to the write up:
Smab’s web app will automatically transcribe podcasts, giving listeners a way to scan and search their content.
What’s the method? I learned from the article:
The company takes audio files and generates text files. If those text files are hosted on Smab’s site, a person can click on a word in the transcript and it will take them directly to that part of the recording, because the transcript and the text are synced. In fact, a second program assesses the audio to determine where sentences begin, making it easier to find chunks of audio. Both functions are uneven, but it’s worth noting here that the company is in a very early stage.
There are three challenges for automatic voice to text to indexing from audio and video sources:
First, there is a great deal of content. The computational cost to covert a large chunk of audio data to a searchable form and then offer a reasonably robust search engine is significant.
Second, selectivity requires an editorial policy. Business and government are likely paying customers, but the topics these folks chase change frequently. The risk is that a paying customer will be disappointed and drop the service. Thus, sustainable revenue may be an issue.
Third, indexing podcasts and YouTube is work that Apple handles rather off handedly and YouTube performs as part of its massive investment in the Google search system. The fact that neither of these firms has pushed forward with more sophisticated search systems suggests that market demand may not be significant.
I hope the Smab service becomes available. Worth watching.
Stephen E Arnold, December 21, 2015
Comments
-
- Subscribe to Beyond Search
Feature archive
News archive
-
