Content Taxonomy Talk
Business exec talks about his
taxonomy product and about the market
6 September 2001
GammaWare software incorporates machine-learning technology, which allows a computer to quickly learn a topic.
Now that’s want you want from a computer.
Dr. Yiftach Ravid, VP R&D of GammaSite, explains, "Our algorithms automatically analyze and understand the content of documents and then automatically organize huge amount of textual data into predefined categories.
GammaWare’s strength lies not only in its high precision, but also in the minimal amount of manual effort required to achieve superb results for hierarchical categorization".
Jay Letwat, business development manager, takes our questions:
Can you Position your product in comparison to others in the market?
Overall, GammaSite focuses exclusively on automatic categorization products, whereas other players have a lot of corporate portal infrastructure products in addition to categorization. The categorization of text is something that is not one of their specialties. Automatic categorization is GammaSite's bread and butter product, something that we have a lot of experience in.
We pride ourselves on very high precision. This stems from our use of proprietary machine learning algorithms developed by our scientists. Precision is truly the bottom line for the end user - is the document in the right category. We consistently achieve precision of 90% and above, of course it is dependent on the information that is cataloged. We have gone head to head with many other players in several deals and we always come out on top. We ran a test for Encyclopedia Britannica with four other vendors and we achieved the highest results.
Minimum manual labor involved is a key point for our customers - how much time does it take to build a taxonomy to achieve the high precision? - For our solution, a few example documents are needed to be input into the system for every category in order for the system to be trained to automatically categorize documents. For other solutions, you need roughly between 50-100 documents per category, which is between 10-20 times more manual work than ours. Imagine for every category having a librarian give 50-100 sample documents. This is extremely time consuming. Many of our customers find this to be the most important aspect of our solution. We can allow them to build a medium sized taxonomy with 200-300 categories in 2 weeks plus 4 weeks for integration and taxonomy building, versus 6-9 months with other solutions.
After a taxonomy is built, there are inevitably changes that need to be made - i.e. new categories need to be made/deleted, other taxonomies need to be built, etc. It is critical that the solution has the ability to make these changes easily.
GammaSite has a Taxonomy Manager which allows categories to be easily added/edited as knowledge needs change.
Most solutions are quite rigid in this regard.
Please describe for us your logical approach to categorization and distinguish it from others out there
Our solution approach is based on the combination of human intervention with supervised machine learning techniques. We believe that this is the best way to achieve information objectives, while at the same time increasing the chances that the results will be very good. Other use "automatic tree creation" which essentially automatically created a hierarchy based on a document set. We find that these types of techniques are not accurate and require a tremendous amount of maintenance in order to structure the tree as desired by the customer. A major flaw with this approach is the fact that the results of the tree generally don't represent the company's knowledge needs.
What are your business targets
We currently have several customers now including The Daily Telegraph, The Daily Mail Group and Xyleme.
We are focusing on 3 major markets including Pharmaceuticals, Financial Services and Publishers/Content
Providers, with a focus on direct sales and OEM' partnerships.
These industries are the most knowledge intensive and are the early adopters of solutions like ours.
While most of our emphasis has been selling to Europe, we are now opening up an office in the US to support current customers as well as further penetrate the US market, critical to our overall company strategy.
The solution is modular, and prices range from 75k to 250k
usd, depending on complexity and amount of licenses.
www.gammasite.com
Delicious
Digg
Reddit
Magnoliacom
Newsvine
Furl
Google
Yahoo
Technorati
Comments
Post new comment