Scott Brown Opfine: "Real-time" financial, company and market sentiment: Part 2
Jinfo Blog

21st February 2012

By Scott Brown

Abstract

In Part 1 of this article, I provided a general overview of Opfine (http://www.opfine.com) and its features. In this part, I look at product highlights and sentiment capabilities.

Item

In Part 1 of this article, I provided a general overview of Opfine (http://www.opfine.com) and its features. In this part, I look at product highlights and sentiment capabilities.

Highlights

In reviewing several articles across company and market results in Opfine, I uncovered a wide variety of sources. For IBM, for example, I found articles from the main sources listed in Part 1, as well as articles from Bloomberg Businessweek, Voice of America, Fox Business, MarketWatch (from the Wall Street Journal), International Business Times and several major US newspapers.

Figure 2: IBM sentiment in Opfine

The topics in Opfine seem to be tied to the appropriate sources. For example, the country category China draws results from Business China, New Tang Dynasty (NTD) Television and ChannelNewsAsia. This doesn’t mean that sources such as Bloomberg Businessweek were excluded, but specialty sources seem to be included where appropriate. In another example, using the company Boeing, article sources included AirlineReporter.com and the Seattle Times, both appropriate to Boeing’s business and location.

At the same time, coverage seems to be a bit uneven. While the IBM example provides a strong sentiment history, in looking at McDonalds – a company like IBM, in that it is well-covered in the media and has a long company history – the sentiment graph seems a bit wonky.

Figure 3: McDonalds sentiment in Opfine

Sentiment

To its credit, Opfine doesn’t oversimplify the task of determining sentiment, and provides at least partial information on its methodology. Once you sign up for an account, you can access a page that provides some detail at http://www.opfine.com/pro/help.jsp.

Essentially, a sentiment score is created for each article, and users see this score next to each article. Once there are enough articles collected, Opfine creates a statistical score assigned to the overall sentiment on that topic for the day. Medla notes that Opfine is a statistical sentiment engine, so it recognises significant events only when there are enough articles in the news sources covering a particular topic.

Opfine also takes into account the fact that sentiment runs on a scale. In other words, not all news items are decidedly positive or negative. The site states, “anything that creates a score above 50 is considered strong sentiment.”

For any sentiment analysis tool, the big question is, how accurate is it?

In my review of several articles rated with strong positive and negative sentiment – across the Macro view, a few companies, and a few countries – I found that Opfine overall did a relatively good job of determining positive and negative sentiment, even with some pretty tricky articles. Compared to a lot of free sentiment tools currently available, Opfine does a decent job of determining sentiment accurately.

Final verdict

Ultimately, how would one use Opfine? Opfine currently provides a pre-constructed base of sources and news for tracking companies, industries and countries. This base, combined with sentiment indicators, could prove very useful as an ongoing information feed. However, the feed functionality is still forthcoming – and sorely needed to make Opfine a more useful product.

The verdict on sentiment is, as usual, a little dicey. I always recommend taking a second look at the sentiment determination for your topic, so you can determine your comfort level with the sentiment accuracy.

Overall, I give Opfine a tentative thumbs up. It’s definitely worth checking out, and I think Opfine will be worth watching over time.

Contact information

Queries can be directed to Marian Medla at marianmedla@gmail.com.

« Blog