Exploring Free Tools for Social Media Sentiment Analysis in Healthcare

flickr: Haiko

I remember the old days of media monitoring (notice I did not say the good old days) when sentiment was measured through press clippings.

As archaic as it seems today, one of my jobs as an undergraduate college intern was sorting and measuring the monthly press clips that were delivered in a large envelope from the university’s clipping service. I would measure the column inches and decide if the story was positive, negative or neutral. I added up the column inches by sentiment to provide the director with a “sentiment analysis” on the coverage of each story.

Today, companies have expanded their monitoring to online conversations and replaced those interns with sophisticated media monitoring and analysis tools.

Assessing online sentiment can reveal patient attitudes about healthcare

Researchers use micro-blogging sites, such as Twitter, to predict sentiment regarding just about everything–from movie box office revenues, to the presidential election, to public health. Assessing online sentiment has also piqued the interest of those in healthcare.

John W. Sharp wrote about the increased interest on eHealth. Being informed and aware of patient sentiment surrounding all aspects of care­­–perceptions about certain drugs, diseases, treatments, specific procedures and particular care facilities­–helps healthcare providers better understand patients and address their concerns (whether perceived or real).

Those of us who engage in sentiment analysis must recognize the wide gaps in accuracy. Even those involved daily in the pursuit, like members of the Association of Internet Researchers, would admit that their work is fluid. It’s evolving, as are the dynamic online conversations and the algorithms created to analyze them.

Even the most sophisticated sentiment analysis tools fall short

Although insights can be gained from some of the tools that exist today, it’s important to remember that those emerging tools only provide a small glimpse of a much larger picture regarding sentiment. Even the most sophisticated sentiment analysis tools cannot fully understand the nuance of human language.

For example, when I used a sentiment analysis tool to analyze the topic “Mayo Clinic,” the results labeled the following tweet negative, when it is clearly positive:  “Every time one of my professors mentions a study by The Mayo Clinic I feel a surge of pride.”

However, with such easy-to-access tools, it is tempting to give them a try. There are a multitude of free tools that have surfaced to analyze sentiment, including Twendz, Twitrratr and SocialMention, and one developed by three graduate students at Stanford called Sentiment140.

Five precautionary steps to sentiment analysis

If you decide to use these early sentiment analysis tools, take a few precautionary steps:

  • Read each tool’s approach to sentiment analysis to better understand the tool’s particular usefulness, as well as the limitations.
  • Use multiple tools and compare the various outcomes.
  • Be aware of current events that could impact results (like the Supreme Court ruling on health reform).
  • Be transparent about the potential inaccuracies when presenting results.
  • Combine online analytic tools with offline tools such as focus groups, patient satisfaction surveys, and staff input to capture a fuller picture of patient sentiment.

Only by being aware and honest about the limitations of the tools we use to measure sentiment, can we provide credible and useful sentiment analysis.


How we help

Hive Strategies gives webinars, presentations and workshops to help hospitals and physician clinics engage patients through social media. Read about our services. Start a conversation. Email us or call us at 503-472-5512.

1 reply
  1. Rob Potschka
    Rob Potschka says:

    Another super tool for Sentiment Analysis but with a twist…Many Online Natural Language Processing services offer an API which is fine for organizations that have in-house developer that can consume the result. Even the organization I work for which has 60k employees; would take a minimum of 2 years before they could approve and build a tool to consume an API based service… So I have built an online tool (looks like the first of its kind) where a company can upload unstructured texts for Online Natural Language Processing while you wait. Its a tool for all those smaller and less agile companies. It’s very fast, uses OpenNLP for Sentence Parsing, Tokenization, N-Grams and Part of Speech combinations to identify Comparative, Personal and Spatial texts in a sentence. All this is visible in a Sankey Flow Report for visual online analysis (I am the developer behind SankeyBuilder.com as seen on Wikipedia). And the results (data linked to sentence/paragraphs) are available for download for offline analysis. The tool has a customized Sentiment Model based on a domain free file (SentiWordNet3) and works very well, better than other commercial tools costing way more. Visit http://sentimentbuilder.com to signup for the always free account! 

    Reply

Leave a Reply

Want to join the discussion?
Feel free to contribute!

Leave a Reply

Your email address will not be published. Required fields are marked *