Author

Topic: Sentiment analysis using bitcointalk thread (Read 524 times)

sr. member
Activity: 466
Merit: 250
I am working on a project that processes bitcointalk's thread, collects data such as posts, members who posted them, their activity, trust etc and then assigns a score to everything, then does sentiment analysis on the post content and tells if the thread is in positive condition or in a bad condition.

The primary objective is to measure the future of altcoins depending on their threads as their threads are the place where the real stuff appears. Currently, I am in the data collection phase.

Has anyone does this before? What do you think about it? How were the results? Looking for productive feedback.
For the majority of the first part (bold texts), yes (achow101 has done it), although it's more in general: https://www.bctalkaccountpricer.info/
Ann thread: https://bitcointalksearch.org/topic/bitcointalk-account-price-estimator-1142314

For the second part, that's something new but there's one issue. In altcoin threads (most of them), there's a huge number of shill accounts that try to make a coin look better and more valuable than they actually are. What process you'll be using to eliminate such conditions for the analysis part?

That project looks really nice. Currently, I havent given much thought to it. The major classification for giving out scores to posters is somewhat similar to what achow101 has done. What I am going for is to check if the project is in a healthy state rather than how valuable it is. Regarding the shill accounts, I will try to figure out a way to clear them out during the data cleaning stage but I havent thought about it yet.
legendary
Activity: 2968
Merit: 3406
Crypto Swap Exchange
I am working on a project that processes bitcointalk's thread, collects data such as posts, members who posted them, their activity, trust etc and then assigns a score to everything, then does sentiment analysis on the post content and tells if the thread is in positive condition or in a bad condition.

The primary objective is to measure the future of altcoins depending on their threads as their threads are the place where the real stuff appears. Currently, I am in the data collection phase.

Has anyone does this before? What do you think about it? How were the results? Looking for productive feedback.
For the majority of the first part (bold texts), yes (achow101 has done it), although it's more in general: https://www.bctalkaccountpricer.info/
Ann thread: https://bitcointalksearch.org/topic/bitcointalk-account-price-estimator-1142314

For the second part, that's something new but there's one issue. In altcoin threads (most of them), there's a huge number of shill accounts that try to make a coin look better and more valuable than they actually are. What process you'll be using to eliminate such conditions for the analysis part?
legendary
Activity: 2940
Merit: 1865
...

That's a very interesting idea.

I once tried to do something similar with the much larger zerohedge.com forum/website/blog.  It is mainly financial, but extremely popular among money-people.  

My idea was to look at the threads there, the general nature of any predictions, popularity etc. with the idea of seeing if there was any PREDICTIVE VALUE (stock market changes for example).  My project turned out to be way too big.  I abandoned it in a week.  Data collection was a problem as they run some 20 threads per day there, and the threads do not "stick around", very fast moving.

Good luck!  Keep us up to date if you go forward.  Such a project looking hard at "Alts" would probably interest many people.
sr. member
Activity: 466
Merit: 250
Hello eveyone,

I am working on a project that processes bitcointalk's thread, collects data such as posts, members who posted them, their activity, trust etc and then assigns a score to everything, then does sentiment analysis on the post content and tells if the thread is in positive condition or in a bad condition.

The primary objective is to measure the future of altcoins depending on their threads as their threads are the place where the real stuff appears. Currently, I am in the data collection phase.

Has anyone does this before? What do you think about it? How were the results? Looking for productive feedback.


Note: If you like what I am doing, you can donate some btc here : 3E1vJTdKqBPewSepnVNjsoy7MPQonwy8X4
Jump to: