Watch this Ask the Expert session to learn how to use SAS Visual Text Analytics to evaluate large, human-generated unstructured text data sources.
Watch the Webinar
You will learn how to:
- Use the right combination of natural language processing, machine learning, deep learning methods and linguistic rules to sift through growing volumes of text data, identify main ideas or topics, extract key terms and analyze sentiment.
- Use human subject matter expertise combined with automatic rule generation and an interactive sandbox to refine results. This will allow you to evaluate subsets of rules to determine which ones are performing better.
- Use the drag-and-drop functionality, best practice templates, simple merge and split features, effortless topic promotion, automatic rule generation and one-click model deployment to reduce the effort required to build models and create more time to focus on finding the information that matters.
The questions from the Q&A segment held at the end of the webinar are listed below and the slides from the webinar are attached.
Q&A
How hard is it to switch from SAS Text Miner to SAS Visual Text Analytics to analyze the unstructured text data?
Since you already know how to use Text Miner, it will be very easy to switch. It's a different interface, but it is quite easy.
Why can’t I analyze my text data using the standard modeling procedure?
For quantitative data, we can have something like 101,515.1 and it will be the same in any languages. But text data we are talking about depends on how each language interpret these things differently. There are so many factors, so many complications, which means we must process all these things differently. That's why we have this special software for analyzing unstructured data that depends on what language you're talking about, how to extract these things and remove the unwanted terms and so on. There is a lot of preprocessing that needs to be done.
Can SAS Visual Text Analytics be used to review a text document for quality assurance?
Definitely. If you don’t have a lot of text documents, you likely won't spend time on it. You're talking about millions, or thousands, of documents and you want to review for quality assurance purposes. If you think these other things need to be there, you need to identify them and look for certain keywords, terms, and so on. You also need create a process of how you want to assess them for the quality assurance purposes. Based on that, you can identify the appropriate rules and categorize them based on whether they satisfy the requirement. You can see how well each document passes the quality check.
For your diabetes example, how could you analyze a clinical visit note for diabetes and rule out people where the note indicates diabetes was ruled out. The word might still be in the note, but it would not be a diabetic.
When you go through the document, you can create specific rules, constraints, or conditions for extracting. This way you'll be creating concept entity with specific restrictions implied on it. When you promote that concept rolled into category, you can identify specific documents related to your conditions. Rather than simply looking for diabetes, you can also add additional restrictions when you extract these concept entities, then promote them to category. That way you can limit what you want to see there.
Is there a tool within SAS to scrape a PDF or Word document to get the data into SAS for text analysis? Some folks use Python to scrape, then import into SAS, but would be nice to scrape using a SAS tool.
We have a tool available within the Visual Text Analytics. If you already have the PDF files and Word files in a folder as individual documents, you can import them into Visual Text Analytics and create SAS data files. We show that also in our training how that can be done.
How much expertise does one need in writing regexp / regular expressions to create custom searches? Your dosage regular expression, for example - did you write that or was it generated by the tool?
We wrote it. In our training, we spend about one day/one session on how to write the LED rules and we explain with examples and demos.
Is SAS VTA only available in SAS Viya?
At present, yes, and we used to have it previously called SAS Contextual Analysis. In the past we had something called Text Miner. If you are using foundation, it's not full-blown visual analytics. It's only for mining text data that can be used to build predictive models. This mining option is still available if you're only interested in building models. But if you are interested in extracting entities, if you want to do sentiment, if you want to categorize, and things like that, then you need to go with Visual Text Analytics.
If there is time: any examples of other utilities using text analytics to find patterns in outage reports or equipment failure reports?
Definitely. If you contact tech support or if you do an online search on Visual Text Analytics, there will be a lot of white papers and examples available. Some are also included in the Resources.
Along with SAS Viya, do we need to have SAS Studio to have control on stop words?
SAS Studio comes automatically. I think it's the programming interface, so we use Studio for scoring new data and so on. If you have SAS Viya, SAS Studio is already included.
What is the difference between the SAS Visual Text Analytics vs. the Text Topics feature in the SAS Visual Analytics application?
Visual Analytics has some components of VTA. It’s good for the people who have doesn't have any license with VTA and want to use some simple text analytics related features. If you want to do, for example, the text cloud map, that is also available in Visual Analytics.
Our organization has a need to scrape emails. How would we connect to Outlook? If we can connect, do we have access to the attachments and are we able to scrape those attachments?
I am not familiar with that, but I'm sure that functionality is included. I recommend contacting the SAS Tech Support and see how they can include the Outlook e-mail attachment.
Once you have custom rules set up, can you auto-schedule imports of new data to be run through the rule(s)?
Yes. It's not built-in within VTA. But, using this as wire scheduling and process, you can automate many things in Viya too. I will recommend contacting tech support on how this can be automated.
Do you need to fix typos and misspellings before the concept definition, or do concepts need to include rules to correct typos/misspellings?
There is an option for if you want to correct this automatically that you can enable. That way it will also check your spelling errors and the like.
What are your thoughts on SAS Enterprise Miner for Text Analytics vs SAS Text Analytics?
The Text Enterprise Miner for is for building predictive models for quantitative data. An add-on to the Enterprise Miner is the Text Miner. If you use quantitative data, but you also want to incorporate text data, then we provided this. But, again, the overall goal is building predictive models. In Visual Analytics, we are not just building predictive models, you can extract information entities text and categorize them. We wanted to find the sentiments during a political debate and see how each candidate is performing live and things like that. There are a lot of useful things needed based on the text data, so VTA is specialized, and the functionality is much more advanced and powerful.
If we want to start using SAS Visual Text Analytics, do we need to get our SAS Admin to install the application? It comes with the Viya package free, correct?
I don't know whether it's free with the Viya package. There may be a trial option. But please check it out with your Customer Success Manager or Account Manager.
Recommended Resources
Get to Know SAS Visual text Analytics – Snap shot demo
SAS is a Leader in The Forrester Wave™: AI-Based Text Analytics Platforms (People Focused), Q2 2020
SAS Visual Text Analytics Demo
Making Text Analytics More Approachable: An Example Using a Customer Service Application
SAS Visual Text Analytics fact sheet
Moving from SAS9 to SAS Viya
Please see additional resources in the attached slide deck.
Want more tips? Be sure to subscribe to the Ask the Expert board to receive follow up Q&A, slides and recordings from other SAS Ask the Expert webinars.
Click image to register for webinar