Searching inside podcasts

Podcasting is bringing attention to the problem of speech-to-text. I have had experience lately with both manual and automated solutions.

1) Podscope and Podzinger are two free services that enable people to find podcasts by the spoken word. My organic chemistry class CHEM 241 was indexed by Podzinger recently and I have had a chance to evaluate its usefulness. That is actually a little difficult because it is not possible to obtain transcripts of the entire podcast through these services. That is understandable because, from the snippet of text displayed in searches, the transcript would be unreadable. I currently view this type of automated speech-to-text as a marginally useful addition to the promotion of any podcast. I have only gotten one hit through Podzinger so far.

2) One of the lectures of my CHEM 243 class was recently transcribed manually by CastingWords. As I mentioned previously, they have chemists on their team and I was impressed with the accuracy. It was not perfect but it only took a few minutes to clean it up so that I could post it on a blog. That way it will be indexed by Google and people will be taken to the podcast from a link in each transcript post. CastingWords charges about $25/hour but I am discussing an advertising model with them also. Hopefully we can work something out to get the rest of my lectures transcribed.

Speech-to-text automation has a long way to go. With the rise of podcasting I see a lot of potential for manual transcription companies in the near future.