Demo HN: Natural Language Twitter Experiment. Beta2 released. | Hacker News Reader

Demo HN: Natural Language Twitter Experiment. Beta2 released.

beta2.tweetspeedreader.com

4 pointscjus6 comments

Pretty cool for the beta.

You can tell the video is made by a technical person talking about 'redirects' and explaining the login process, and describing the minutia of how each process works. If they are looking to go mainstream with this, I'd recommend a more dynamic demo with less description and more about how useful it is.

At the same time, I don't find the UI very compelling. I was expecting a 'timeline' showing tweets with similar themes, etc. This doesn't seem much different from reading my tweets. By gathering some metrics on the number of followers from the tweeter (reader being tweetee) TSR may be able to provide a better method of showing you the most popular tweet on a given topic. Tag cloud navigation doesn't strike me as effective use of NLP vs theme clustering. But maybe that's just me.

Best of luck

pedalpete, excellent feedback - thank you. Clustering doesn't seem to be conducive to the benefits of the long tail. Wouldn't clustering simply show you what's popular? Do you happen to have links discussing theme clustering? A quick Google search didn't yield anything tangible.

I meant to get back to you after Beta1... The product is interesting, but wouldn't sway me from using a normal twitter client. For my use of Twitter I care more about who is saying what than what was said overall, if that makes sense.

You might want to refocus it as a trend or search type app. If you can out trend twitter trends by picking up on context and not just tags, I think you'd really have something.

Crad, thanks for your time. My goal isn't to replace twitter clients - but to offer a place one checks from time to time to get caught-up and possibly discover interesting tweets which were simply lost in the noise.

BTW, do you rely on the use of "lists" to keep up with "who is saying what"? Or do you simple resign that to a serendipitous event? I think this is what most people do. Another problem I'm trying to solve...

I find this "experiment" interesting because doing it well has widespread application and benefits.

I mainly use lists to export the people I follow into carefully packaged categories. I usually just tune in and tune out to the stream and miss vast amounts of tweets.

I absolutely love it (seriously). I can honestly see myself using this very often. Good job to the team!

Reply on news.ycombinator.com