Product description: The DiscoverText program can import, sort, distribute, and analyze electronic content from email, document repositories, and social media. Textifers’ describes the future of text analysis as a need for a unified repository for the “eDiscovery” process. The underlying idea is to close the loop between humans and machines in data analysis.
This program applies crowdsourcing to text analysis. DiscoverText combines both software algorithms along with human-based coding to provide accurate analysis. Originally developed to help federal agencies sort huge amounts of public comments, the program is being tailored for appropriate scale and sharing.
DiscoverText is a cloud-based social media aggregator with a search engine behind it. They host your data and analysis on their site so you don’t install the program on your computer. The creators aim for a hybrid where you create your coding and use thier global classifiers. As well as bridging the gap between human and machine, this system aims to be a peer based system for sharing coding, datasets, and information.
This webinar introduces the beta version of training sentiment detection and mood classifiers. Since this is a beta version, I got a real feel for the training process for classifiers. I appreciated this “behind the scenes” view of the challenges of sentiment analysis. A case sample of Tweets about AdWeek mispelling Zynga as Zenga pointed out issues of interpretation of positive words that make for negative (cynical) comments. Sentiment analysis is at the stage of tackling the subtleties of human language that confound and confuse trained classifiers. The de-duplication system sets a threshold for similarity that clusters and displays Re-Tweets with slight variations.
The 90 minute evening tutorial was shared with a group of Korean business researchers who were in a morning class-being on the other side of the world. The questions and comments were knowledgeable, interested, and direct.
The creators are very anxious for people to jump in and test the entire system, as well as the beta sentiment classifer system. For the beta system email Stuart Shuman at firstname.lastname@example.org.
There is a free Community edition that has some great functions and tools. You will learn alot about this type of analysis by just playing around with the tools and the sample datasets. Jump on in!
Look at News and Updates for short infotorials and upcoming webinars, a short video on scraping info off of Facebook, and harvesting Tweets through hashtag searches.
A free 30 day trial is available. Learn more at: http://discovertext.com/defaultDT2.aspx