FACT SHEET

SAS® Visual Text Analytics

Combine the power of natural language processing, machine learning
and linguistic rules to reveal insights in data

What does SAS® Visual Text Analytics do?

SAS Visual Text Analytics offers a wide variety of modeling approaches for getting the most
value from unstructured data, including supervised and unsupervised machine learning,
linguistic rules, categorization, entity extraction, sentiment analysis and topic detection.

Why is SAS® Visual Text Analytics important?

Text data is pervasive across all industries with volumes growing each day. SAS Visual Text
Analytics provides a comprehensive solution that overcomes the challenges of identifying
and categorizing text data. It enables organizations to scale the human act of reading,
organizing and extracting useful information from huge volumes of textual data. You build
models that analyze and categorize a set of documents. Unstructured data is automatically
converted into meaningful insights that feed machine learning models.

For whom is SAS® Visual Text Analytics designed?

It’s designed for business analysts, domain experts, research analysts, linguists, knowledge
workers and data scientists who need to analyze large amounts of unstructured data to
glean new insights.

Text analytics helps Benefits

solve a variety of
• Uncover emerging trends and spot new device, that data is analyzed immediately
everyday business
opportunities for action with artificial and accelerates the data-to-decision
problems – things
intelligence. Automatically convert timeline. Organizations can reduce the
like managing
unstructured data into meaningful gap between when information is received
and interpreting notes, assessing risk or
insights. Increase the accuracy of text and when it’s acted upon with in-memory,
fraud, and incorporating customer feedback
models by combining machine learning in-database and streaming technologies.
for earlier detection of problems. With more
unstructured data than ever, the use of text methods with a rules-based approach • Fuel collaboration and information
analytics is expanding across all industries. that can be enhanced with subject- sharing in an open ecosystem.
matter expertise. SAS Visual Text Analytics provides a
SAS Visual Text Analytics analyzes large • Automate comprehensive analytics. flexible environment that supports the
volumes of unstructured data using SAS Visual Text Analytics offers great entire analytical life cycle – from data prep-
predefined templates, machine learning breadth and depth of analytical capabilities aration and visual exploration to analysis
methods and natural language processing through a rich mix of rules and machine and deployment. Tackle and experiment
(NLP) to produce deeper insights using learning, as well as integrated deep with a variety of analytical use cases to
more data, faster than ever before. This learning. It provides quick start pipelines support a single initiative. Whether you’re
software combines text mining, contextual and can use industry taxonomies to readily a data scientist preparing data, a domain
extraction, categorization, sentiment analysis support both predictive and prescriptive expert applying linguistic rules or an
and search within a modern and flexible analytics. IT person deploying models, collabora-
framework. An end-to-end visual pipeline • Go from data to decisions more quickly. tion is possible at all levels. This unified
makes it easy to prepare data, visually Empower decision making at the source solution integrates seamlessly with
explore topics, extract entities and facts, of the data. If someone leaves a comment existing systems and open source
analyze sentiment, build text models and or clicks through an app on a mobile technologies.
deploy them within existing systems or
processes.
Overview range of Boolean operators and linguistic SAS Visual Text Analytics provides the most
comprehensive set of tools for achieving
qualifiers, users can dig deep into their
When the volume of text-based data becomes precision and contextual specificity to aid in
unstructured data.
too large to manually review and analyze, contextual extraction with a wide range of
it’s time to add text analytics to your insights
Contextual extraction rule types, Boolean and linguistic operators,
arsenal. SAS Visual Text Analytics supports term qualifiers, regular expressions, part of
the entire analytics life cycle with data manage- The contextual extraction technique is often
speech tags and more.
ment, comprehensive analytics and flexible used to isolate and pull out important pieces
deployment options. of information where the value of matches
to specific context is of utmost importance. Flexible deployment
Embedded data preparation By extracting snippets buried in free-form With SAS, you can deploy models in batch,
and visualization text without the need for manual markup, in Hadoop, in stream and via APIs. Data
SAS Visual Text Analytics comes with you can derive new variables to incorporate does not have to go through the user inter-
embedded data integration and prepara- into reports, predictive models, or enhance face to be enriched by a text model. Models
tion capabilities to help access, integrate, search and filtering applications. You can can be run closer to where data is located.
profile, cleanse and transform data. You can use predefined concepts to detect and This reduces data movement and produces
import text directly from more than 35 out- extract data elements and relationships from faster results for scoring new data.
of-the-box data connectors, including unstructured text, or you can create custom
multiple document formats, relational data- concepts and definitions.
bases, remote file system data sources, local
data file types, social media connectors and
Esri. The software also includes self-service
data visualization capabilities for exploring
and understanding your text data.

Combined machine learning,

deep learning and rules-based
methods
SAS Visual Text Analytics combines a rich
mix of artificial intelligence (AI) techniques
(supervised and unsupervised machine
learning and deep learning) and rules-
based approaches to automatically surface
themes in text data, generate rules for cate-
gories of interest, produce visual representa-
tions of related terms for exploratory analysis Automatic topic discovery groups documents based on common themes.
and create best practice templates for
specific business use cases.

SAS Visual Text Analytics derives topics from

groups of important terms in your docu-
ments. You can explore trending topics in
text and see how they change over time.
These insights can kick off immediate
actions. Or, you can fine-tune them with
subject-matter expertise.

If discovered themes require additional

tuning or custom definitions are preferred,
you can create precise, rules-based catego-
ries and concepts. Despite its advances,
machine learning still can’t capture specific
nuances and complexities of language ambi-
guity. With advanced rule tuning and a wide
Powerful NLP capabilities allow you to easily identify and extract important patterns in text.
Multiuser environment fosters Key Features
teamwork and collaboration
Machine learning and rules-based approaches in a single project
As part of the SAS Platform, SAS Visual Text
• Unsupervised machine learning automates topic generation.
Analytics fosters teamwork by providing a
• Supervised/probabilistic machine learning models include BoolRule (enables auto-
workspace to share best-practice pipelines
matic rule generation for document categorization), and conditional random fields
and methods. Any extensive analysis, such and probabilistic semantics (used to label and sequence data and can automate
as identifying relevant terms, modifying or entity and relationship extraction by learning the contextual rules of a given entity).
creating user topics, creating linguistic rules, • Automatic rule builders promote topics to categories with supervised machine
etc., can be shared as a Best Practice node learning.
in a pipeline across multiple SAS projects. • Rules-based linguistic methods are used for extracting key concepts.
• Rules can be tested on an input data set prior to deployment.
SAS also integrates with existing systems • Automatic parsing can be used along with deep learning algorithms (recurrent neural
and open source languages, including networks) to classify documents and sentiment more accurately.
Python, R, Java and Lua. And with RESTful
APIs, you can easily add the power of SAS
Contextual extraction
to other applications.
• Use predefined concepts to extract common entities such as names, dates, currency
values, measurements, people, places and more.
The software provides an intuitive user inter- • Create custom concepts using keywords, Boolean operators, regular expressions,
face that accounts for important factors such predicate logic and a wide array of linguistic operators.
as localization/internationalization and acces- • Programmatically populate fields that contain the desired matched information.
sibility. Additionally, by offering open APIs No need to manually mark up document for entities or facts of interest.
and a microservices architecture, SAS offers • Reference a predefined or custom concept in a categorization rule, for extra contextual
users the ability to use their own user inter- specificity or reach.
face or build a custom search application.

Flexible deployment
Natural language processing
• Concepts, Sentiment, Topics and Categories nodes provide score code needed to
Manually reviewing documents is time- deploy models on an external data set.
consuming and prone to errors. Natural • Score code is natively threaded for distributed processing, taking maximum advan-
language processing (NLP) reduces the tage of computing resources to reduce latency to results, even on very large data sets.
need for tedious manual analysis and allows • Register text models and monitor performance and model decay with SAS Model
you to more easily identify and extract Manager.
important patterns in the text.

NLP is often the first step in the text analytics

Multiuser environment
• Graphical user interface with visual programming flow.
process. It performs linguistic analysis to
• Share projects with other users.
help a machine “read” text. SAS Visual Text
• Five nodes available through text analytics pipeline (Concepts, Text Parsing,
Analytics uses NLP to analyze and transform
Sentiment, Topics and Categories).
text into formal representations for text • Prepopulated default pipeline to represent typical workflow of a text analytics project.
processing and understanding. • Flexible pipelines enable users to create additional nodes or modify default pipeline.
• Register models in SAS Model Manager for easier management of text models.
There’s no need to cobble together dispa-
rate NLP libraries or custom code to
perform word and sentence tokenization, Natural language processing
segmentation, stemming, compound • Automated parsing, tokenization, part-of-speech tagging and lemmatization.
decomposition, part-of-speech tagging, • Ability to apply start and stop lists.
• Ability to use special tags, qualifiers and operators in linguistic rules that take
named entity recognition and semantic
advantage of part-of-speech tagging, tokenization and lemmatization (allows for
parsing. You spend less time programming
more precision or better recall/abstraction capabilities).
the computer how to interpret text and
• Detect misspellings.
more time deriving business value from
textual data.
Automated feature extraction with machine-generated topics
• Automatic topic discovery groups documents based on common themes; each
document may contain zero, one or more themes.

Continued on reverse
Automated, machine-generated Key Features (continued)
topic detection
SAS Visual Text Analytics provides two • Relevance scores are produced that characterize how well each document belongs
to each topic, as well as a binary flag showing topic membership above a given
machine learning methods for automatic
threshold.
topic discovery within your documents:
• Merge or split topics automatically generated by the machine (unsupervised machine
singular value decomposition and latent
learning) to create user-defined topics (subject-matter expertise to refine automated AI
Dirichlet allocation, which is popular within
output).
the open source community. With these
methods, you can quickly discover text topics
and inspect the terms and documents that Native linguistic support for multiple languages
make up these natural groupings. Discover • Out-of-the-box text analysis for 32 languages: Arabic, Chinese, Croatian, Czech, Danish,
themes you may not have thought to look Dutch, English, Farsi, Finnish, French, German, Greek, Hebrew, Hindi, Hungarian, Indone-
for. Then, deploy the model against new sian, Italian, Japanese, Korean, Norwegian, Polish, Portuguese, Romanian, Russian, Slovak,
data or convert topics to logic that can be Slovene, Spanish, Swedish, Tagalog, Turkish, Thai and Vietnamese.
manually modified or extended by subject- • Default stop list provided for each language the application supports.
matter experts. • Built-in lexicons that contain part-of-speech information and dictionary-based expan-
sion to detect and resolve surface forms to root form (verb conjugations, plurals, etc.).
Native linguistic support for
multiple languages Sentiment analysis
SAS Visual Text Analytics supports a wide • Visual depiction of document-level sentiment through sentiment indicator display
variety of languages through dictionaries at a document and topic level.
and linguistic assets created by native • Default domain-independent sentiment analysis taxonomy for 14 languages: Arabic,
language experts. This helps support the Chinese (simplified), Chinese (traditional), Dutch, English, Farsi, French, German,
global challenges organizations face. Italian, Japanese, Korean, Portuguese, Spanish and Turkish.
Standardization of part-of-speech tags and • Ability to import and execute custom sentiment models built in SAS Sentiment
named entities across languages is key for Analysis.
organizations looking to implement text • Ability to use recurrent neural networks for more accurate sentiment classification.
analytics in a variety of languages. SAS Visual
Text Analytics includes out-of-the-box Open APIs
analysis functionality for 32 languages. • Seamlessly integrate with existing systems and open source technology.
These language packs enable native language • Add the power of SAS Analytics to other applications using SAS® Viya® REST APIs.
analysis as opposed to language translation • Out-of-the-box analytical programming interfaces for text summarization, text data
prior to analysis. segmentation, text parsing and mining, topic modeling, text rule development
and scoring, text rule discovery, term mapping and topic term mapping, conditional
Sentiment analysis random field and search.
Sentiment analysis identifies an author’s tone
or attitude (positive, negative or neutral) that
is expressed through text. SAS Visual Text
Analytics identifies and analyzes terms,
  TO LEARN MORE  » 
phrases and character strings that imply
sentiment. Continuously track sentiment to To learn more about SAS Visual Text
see how dimensions of interest are changing. Analytics, view screenshots and see
Better understand and categorize feedback other related materials, please visit
and adjust decisions based on perspectives. sas.com/vta.

To contact your local SAS office, please visit: sas.com/offices

SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc.
in the USA and other countries. ® indicates USA registration. Other brand and product names are trademarks of their
respective companies. Copyright © 2018, SAS Institute Inc. All rights reserved. 109227_G80168.0718