Sentiment Analysis: The power of the crowd

We live in an age of too much information and yet, we do not have access to the right information, at the right time. An accurate quantification of the sentiment of relevant text that is published online (referred to here as the sentiment score), can help solve this daunting problem.

There are two main aspects that we will be examining in this article: what types of problems can sentiment analysis help solve, and how complicated is sentiment analysis?
 

Consumer Sentiment: Revealing the Facts

Sentiment analysis of open online sources now provides the most accurate view of customer feedback and satisfaction.

More importantly, sentiment analysis over time can show investors which direction a particular company is heading.

For example, after the release of Apple’s iPhone 4s, the media perception of the device was by and large negative.

However an analysis of twitter sentiment for the phone revealed a positive consumer trend, and after 3 months it was declared to be a blockbuster product, with the associated affect on the share price.
Consumer sentiment is the only way to differentiate facts from perception.
 

However sentiment analysis is not only about looking at sentiment direction (positive / negative).

In another incident, there was general negative sentiment on
twitter following the recall of a car model from Toyota in 2009, with people across different regions complaining of issues with the car.

But when a respected person spoke negatively about it, there was a sudden drop in the share price.

This shows that the intensity of impact will also be affected by the reliability and standing of the source and weighting associated to it.
 

Dynamics of Sentiment Quantification

When looking for a sentiment score, particularly in finance, where every event is connected to multiple incidences, the whole ecosystem acts as a contributor.

For example when a drug is recalled by a Pharma company, the algorithm to quantify the sentiment of that news will look up multiple aspects, such as if the drug was patented or generic, as a patented drug will have a much higher impact on the share price, as compared with the generic drug.

Also is there any recent news about the company, such as has any other drug trial recently failed or has the FDA commented negatively on other trials, as well as many other aspects.

Sentiment scoring algorithms go much further than just looking at the volume of keywords present, they consider the entire ecosystem before
quantifying and scoring it.
 

Using Sentiments or Fundamental and Technical Analysis

Sentiment is a very versatile tool, which can add new dimensions to both fundamental and technical analysis. While doing fundamental analysis, sentiment equips the investor with useful prior knowledge about the sector.

‘Sentiment is a very versatile tool, which can add new dimensions to both Fundamental and Technical analysis. While doing Fundamental Analysis, Sentiment equips the investor with useful prior knowledge about the sector’

Recently a sudden awareness of environmental hazards of palm oil, created deep negative sentiment on social media and blogs for Proctor & Gamble.

The company has now changed its manufacturing procedure for those lines of products, however potential investors could have been forewarned by analysing the trending sentiment of the company in question.

For technical analysis Sentiment time series analysis, can provide a lot of information about the company’s timeline and is a very strong indicator along with volume and moving average.

Companies are now performing analysis to predict trend reversals in share prices, using sentiment direction as indicators along with frequency of tagged words, weighing of sources, volume and price data; which gave results with an accuracy in excess of 70%.

Initial Claims Prediction

The above image shows how twitter sentiment is currently being used to predict US unemployment claims, a major macro economic indicator.
 

By processing millions of tweets and calculating the sentiment of the general public The University of Michigan has started to show a correlation with the official figures being released.

Other companies, such as Heckyl, have taken on these challenges and are refining the process still further to calculate even more accurate forecasts of US Jobless Claims.
 

Sentiment Analysis: The way forward

Many companies have started to adopt Sentiment analysis in taking key decisions and it is now becoming accessible to retail investors also. State of the art algorithms are continuously being researched and implemented to improve the efficiency of Sentiment analysis

These will only get better as they are refined and the number of users and content from the sources, such as twitter and blogs, continue to increase.

We are in middle of an era of information explosion, where assimilating all the information is no longer feasible.

Sentiment quantification is the answer to this and the applications of sentiment analysis are infinite.

Click to visit: