I got an update on release 8.0 from Angoss last year and they subsequently released 8.5 in November 2012 with new functionality in KnowledgeSTUDIO and KnowledgeSEEKER. Particular improvements came in their additive scorecard interface with the addition of reject inference. Data visualization and reporting got an update too and they added the ability to import R datasets. Release 8.5 Specifics:
- For R they allow users to import and export R data files so that data miners and analysts can do preliminary work in R and then bring the files into Angoss to perform visual data discovery and model comparisons with Angoss’ Decision Trees and interactive modeling features
- Scorecards and reject interface added support for a weight of evidence editor as well as support for Proportional Assignment, Hard Cut-off and Parceling methods for reject inference (essentially ways to infer the characteristics of those who were rejected and so don’t show up in results)
- Text analytics enhancement included extending integration with Lexalytics’ Salience engine by adding Concept Topics for coarse grained classification (Lexalytics builds this by analyzing Wikipedia to find topics) and Query Topics so users can define their own topics. Users can also define their own entities like product lines or product names.
- On the visualization side a process map was added to give a visual trail of the analytical process – all the steps in the project shown in a visual hierarchy that is recorded as you work.
- New reports include cumulative lift and lift reports, decile by decile performance for instance.
- Improved chart management inside the tool plus some new univariate statistics rounded it out.
In April, Angoss is releasing 8.7 with a focus on text analytics. Currently the tool can do entity, theme and topic extraction, visualize relationships, generate datasets from this, push this into their modeling tools and handle things like case conversion.
New features for 8.7 are very focused on visualization of the text analysis results:
- A dashboard with a visual display of sentiment for instance. These are predefined by Angoss and generated automatically, though the user can control some of the rendering and categorization.
- Trend and comparison analysis for sentiment. For instance side by side graphs of sentiment across different product categories, top 10 topics compared across categories etc.
- Sentiment Phrases to model what’s driving sentiment and find words that drive it out of bounds, for instance product names that use words with high negative sentiment.
- Sentiment Category to group into positive, negative, neutral from the more granular approach
- Association discovery between these elements using the market basket algorithm and applying it to the text analytics results, displaying the associations visually as an item attraction map. Again users can control size, color and layout parameters to investigate their results. All of the underlying information is available as a structured data set for analysis.
- Markup to allow the original text showing sentiment and part of speech markup in the text. Users can take topics or phrases that they are investigating and drill down to the original text where the phrases and categories are shown in context. Multiple themes and phrases can be selected to find situations where both are used.
While Angoss sees folks who are just getting started with text analytics using text analysis as an exploratory technique, their more advanced customers are definitely driving towards using text data sources as part of their predictive models.
Angoss is one of the vendors in our Decision Management Systems Platform Technologies Report, and you can get more information on Angoss KnowledgeSEEKER and KnowlegeSTUDIO and by visiting www.angoss.com.