Salience 6 is built on the Syntax Matrix, a powerful new technology that supports cool new features, like extracting intent with our uniquely potent intention analysis, as well as enhancing existing sentiment analysis.
The newest version of Salience also offers new avenues for customization, including the ability to use sets of tagged content to train custom classification models.
Salience 6 is currently available for on premise customers and the new features will be available in our Semantria SaaS text mining system on December 16th.
Understand language structure with the Syntax Matrix
We’ve been working on our brand new Syntax Matrix for over a year now. The deep learning Syntax Matrix is a quick and computationally efficient way for understanding the structure of language. The Syntax Matrix is trained on billions of words in order to understand how the phrases in a sentence relate to each other. This enables you to quickly and efficiently extract meaningful and actionable results.
Take immediate action with Intention Analysis
The Syntax Matrix powers one of Salience’s most powerful new features, intention analysis. Intention analysis, as the name suggests, detects the intent of a customer in text. Other text mining systems use simple keyword analysis to indicate intent based on the presence of a word like “buy”. Thanks to the Syntax Matrix, we can identify intent without being restrained to a keyword, making our intention analysis much more useful and robust.
Out intention analysis works by extracting the intent, the “intendee”, or person voicing the intent, and the object of that intent. For instance, “Because of my unhealthy binge-watching habits, I’m considering quitting Netflix for the sake of my professional career.” In this sentence, the intent is “quit”, the intentee is “I”, and the object of the intent is “Netflix”.
The great thing about Intentions is that it extracts necessary information to make a business decision so you can take immediate action to find leads, develop new revenue streams, and route social media support requests, and more.
Currently, Intentions detects the intent to buy, sell, recommend, or quit. We are looking into other types of intent, and would love to hear feedback about useful intentions to detect.
Achieve greater accuracy with Enhanced Sentiment and Alternative Forms
Salience 6.0 isn’t just about greater swiftness and efficiency, we’ve also features like alternative form detection and enhanced sentiment to improve the accuracy of your results.
Most people don’t have the benefit of a live-in copy-editor to proofread their tweets and status updates, which means they tent to misspell things, especially on social media. As Salience is used to analyze more and more non-journalistic content, we’ve needed to keep with the times and introduce new techniques to deal with this human error. Detecting alternative forms is one of these techniques. Salience does not auto-correct words, but it does detect words that are possible misspellings or slang forms of a word and analyzes them as such, which makes for better results.
We’ve also upgraded our sentiment analysis. The Syntax Matrix makes it much easier to analyze sentences that contain mixed sentiment. For instance, “Tumblr is amazingly engaging and absorbing, but it is so addictive it is literally ruining my life.” The first clause indicates positive sentiment about Tumblr, but the negative sentiment that follows carries more weight and contains the real message of the content – Tumblr’s great and terrible destructive powers.
Salience also handles sentiment expressed in lists and tables now, using table and list headers as cues to the sentiment contained within these structural elements.
Customize better for your needs with Entity Extraction Enhancements and Custom Classification models
We’ve responded to customer requests for greater customizability with Entity Extraction Enhancements and Custom Classification models.
Out of the box entity extraction supports the extraction of People, Places, Companies, and Products. Salience 6 lets you go beyond these with the ability to easily define your own entity types. We’ve also expanded Company extraction to better extract company-like entities such as organizations, associates, political parties, and government agencies.
Finally, we’ve introduced new machine learning classifiers and improved our query grammar so you can create your own custom classification models. This is especially useful for those of you who already have large sets of content that have been tagged manually. Instead of starting from scratch, leverage the all the hard work you’ve already done. Use your pre-existing taxonomy to train classification models to match your pre-existing coding rules. Our customization options ensures that Salience is tailored to your needs.
Give us your feedback
We’re eager to watch our customers start playing with the new features we’ve built, and to start building even more. Our new Syntax Matrix technology in particular provides an excellent new foundation for us to build on in the future. As out customers, your input is crucial. We look forward to your feedback as you begin working with our improved system.