Automatic generation of interesting thing to read?
While I believe in the tech of statistical NLP, I question whether the quality of news articles automatically generated from other news and social media will be all that interesting.
I spend about 3 to 4 hours a week reading through Twitter, HN, and Reddit: half is to waste time in an enjoyable way and half is to find interesting articles to read, new useful projects, etc.
For me, Twitter is the most targeted because I follow people into the same tech that I am into. I also have several blogs I follow closely. A big part of it is enjoying authors' online personalities and having occasional email dialogs.
Can an automated system replace part of this experience? I don't think so.
What automated systems can do is cluster reading material and make good recommendations - but this is different than what (it sounds like) Wavii is trying to do.
Imagine if all your twitter data was tagged by content and author and there was an easy, even automatic way to find new followers based on those tags. Seems like it could add value on top of twitter, given the right number of users.
The real power of an application like this would be aggregating real-time news events as they unfold and creating news stories from real-time tweets, blog posts, and other social media. Instead of just clustering together topics, being able to "read" a large amount of sources and generate a summary would actually be very useful. Unfortunately, I think that it ignores a lot of very difficult problems in natural language processing and computational linguistics. It will likely have a few homeruns, but peter out as it hits an accuracy wall.
5 comments
[ 3.2 ms ] story [ 21.6 ms ] threadWhile I believe in the tech of statistical NLP, I question whether the quality of news articles automatically generated from other news and social media will be all that interesting.
I spend about 3 to 4 hours a week reading through Twitter, HN, and Reddit: half is to waste time in an enjoyable way and half is to find interesting articles to read, new useful projects, etc.
For me, Twitter is the most targeted because I follow people into the same tech that I am into. I also have several blogs I follow closely. A big part of it is enjoying authors' online personalities and having occasional email dialogs.
Can an automated system replace part of this experience? I don't think so.
What automated systems can do is cluster reading material and make good recommendations - but this is different than what (it sounds like) Wavii is trying to do.