5 Key Challenges in Sentiment Analysis

Takeaways from a recent symposium…

Plutchik's Wheel of Emotions EnlargedAs the adoption of sentiment analysis continues to spread across industries, from politics to PR, opinions about the field also run deep. That’s especially true among practitioners, and a range of academic and vendor specialists weighed in at the Sentiment Analysis Symposium in New York last week. While the novelty factor begins to subside, clients are looking for more substance, and as befitting such a multifaceted topic, it’s complicated.

As a follow-up to yesterday’s post that covered the analysis of visual images and facial coding, here the experts offered their perspectives on approaching 5 ongoing issues:

1. Tread carefully on accuracy numbers

The degree of accuracy issue is hard to answer, said Bing Liu, a University of Chicago computer science professor specializing in data mining. It depends on what you’re measuring, the level of text you’re analyzing, the number of data sets across domains and the voice sound quality of videos, among other variables. Still, he thinks that progress is being made in this regard.

Rob Key, CEO of social media consultancy Converseon, offered a prime example of what can potentially go wrong with monitoring social media. He referred to an article from The Atlantic about whether Anne Hathaway news drives Berkshire Hathaway’s stock prices. Key said this is due to some hedge funds using primitive data sets, where Hathaway mentions aren’t properly differentiated.

2. Utliize both machine learning and human knowledge

“Machines do analytics, humans do analysis”, remarked Anjali Lai, an analyst at Forrester Research, as she advocated for a blended approach.

And as professor Liu noted, machine learning is isolated and humans don’t learn in isolation. So it’s critical for humans to apply the prior knowledge they’ve gained from their experience.

Key agreed, saying it’s important to keep humans in the loop for continuous training (and to guard against Hathaway type headaches).

3. Adopt a multi-method research plan

Social sentiment data sometimes doesn’t explain why an event occurred, or among which demographic group, said Lai. And the downside of much survey work is that it only measures respondents’ reactions at one point in time. So Forrester favors conducting both sentiment analysis and surveys to add verbatim comments. That serves to help explain rationales and to match sentiment data to the relevant audience targets.

4. Keep an open mind about the findings

All too often, clients approach sentiment analysis with specific hypotheses in mind, and selectively extract data that proves their theories, Lai lamented.

They’d be better off waiting to check the wealth of information uncovered. They may even be surprised, as Brooke Miller, CTO at tech insights company MotiveQuest, pointed out. One of his clients, Claussen pickles, learned from social media monitoring that some athletes and trainers were using pickle juice to help stop muscle cramps. That turned into a new brand benefit and usage scenario.

5. Stop treating sentiment analysis as a hobby

That was Key’s request, and he’d also like to see more focus on prediction rather than merely retrospection based on the prior data generated.

Stephen Rappaport also sees the need for sentiment analysis to get more serious. He’s the author of the Digital Metrics Field Guide and a senior digital advisor at Sunstar, and often conducts workshops for corporate executives. He’s observed that in many companies, the analytics function isn’t mature enough yet to contribute to business growth. He said a higher level of senior management knowledge as well as increased staffing are necessary in order to take a more sophisticated analytical approach.

(Image: Plutchik’s Wheel of Emotions)