Lexalytics Simplifies and Improves Text Analytics for the Enterprise with New Machine Learning Capabilities

Print Friendly, PDF & Email

Lexalytics_logoLexalytics®, a leader in cloud and on-prem text analytics solutions, announced that it has bolstered the machine learning (ML) capabilities of its Salience text analytics platform, making it easier for data analysts and scientists to train their Salience software to deliver actionable insights from data sources. In addition, Salience 6.2 now enables professionals in social media marketing, voice of the customer (VOC), voice of the employee (VOE) and customer experience management (CEM) to more accurately analyze email communications, as well as text that includes the latest emojis.

Lexalytics is one of the best text analytics engines on the market,” according to Nova Spivack CEO and co-founder at leading cognitive computing company, Bottlenose. “With this latest version of Salience, they’re making it a lot easier to extract insights from unstructured data.”

Salience is Lexalytics’ on-premise text analytics engine that processes billions of documents per day, powering seven of the top 10 social monitoring and social marketing providers, as well as leaders in customer experience management, survey analysis and business intelligence software.
New features of Salience 6.2 include:

  • HSDTrainer — Salience 6.2 features a unique new tool to easily “train” sentiment analysis, combining the ease of training of a simplistic machine learning (ML) system with the transparency of Lexalytics’ natural language processing (NLP) technology. For example, if you were to train solely on content without any view into how the system is making its decisions, that system might learn that the phrase “Greek bank” is negative, due to the deluge of negative stories associated with Greek banks over the years, even though the phrase is not inherently negative. This is a common problem with systems that attempt to analyze sentiment with a single model and will skew results over time. The Lexalytics HSDTrainer can consume any text corpus that has been appropriately marked up for sentiment, and then return a list of phrases and suggested scores for that text corpus, allowing analysts to both rapidly and transparently train sentiment.
  • Emoji Analytics — With Salience 6.2, social marketers can now analyze the meaning and sentiment of content that includes the latest emojis released in Unicode 9.0. For example, if a food manufacturer releases a new product that elicits social media posts with the new “nauseated face” emoji, Lexalytics can score the content as negative and alert the customer. Conversely, those same marketers can search for anything that mentions “nausea,” and that emoji will return a hit.
  • Email Processing — Salience 6.2 adds the ability to ingest email databases, stripping out headers and footers, eliminating duplicate emails and analyzing email threads efficiently. Email processing can help companies detect siloization within their organization or improve the process of analyzing customer support emails to see what customers are particularly vocal about.
  • Improved Named Entity Recognition — Combining the power of Lexalytics’ sophisticated machine learning models and known lists of people, places and things, i.e., “entities,” Lexalytics has improved its precision and recall scores, known as F1 scores, by up to 25 percent. With Lexalytics’ growth in Asian markets including China, Japan and Korea, Salience 6.2 improves upon its ability to more readily recognize people’s names from those regions.

With Salience 6.2, we are dramatically bolstering our ML capabilities, making it easier for our customers to teach and ‘tune’ the software to meet their unique needs,” said Jeff Catlin, CEO, Lexalytics. “Ultimately we’re improving the way machines and humans interact by expanding the capabilities of what text analytics and NLP can accomplish.”


Salience 6.2 from Lexalytics is available now. For more information, please visit https://www.lexalytics.com/salience/server.


Sign up for the free insideBIGDATA newsletter.

Speak Your Mind