Supporting the value of information in your enterprise. How?

Features: Article

All FreePint Articles | Article Categories |

FINIF mini review: Part 2
Wednesday, 7th March 2012 Please login to be able to star items

Abstract:

Article:

In Part 1 of this review, we looked at the basic premise behind FINIF, a financial informatics platform. FINIF is not only unique in that it is one of a very few in the financial sentiment analysis niche (see also the recent Opfine review by Scott Brown – http://web.vivavip.com/go/livewire/66640) but the creators, University of Notre Dame Professor Dr Bill McDonald and his son, FINIF CEO Michael McDonald, have also created a proprietary dictionary of key words that is specific to filings that US companies must make to the Securities and Exchange Commission (SEC).

In short, McDonald discovered that many of the sentiment words used in analysis platforms came from psychological and sociological based dictionaries, such as the Harvard Dictionaries. This often leads to misinterpretation when it comes to business or financial documents. Often, a term that may have a negative connotation in the marketing world, such as “vice”, is used rampantly in a 10-K filing to describe vice-president. Add that to the practice in sentiment analysis of providing a great deal of weight and focus to what are considered negative words, and you have a recipe for misguided financial sentiment analysis.

As a matter of fact, Dr McDonald even co-authored an article in 2009 for the Journal of Finance entitled When is a Liability not a Liability: Textual Analysis, Dictionaries and 10-Ks (http://www.afajof.org/afa/forthcoming/6989p.pdf) (PDF). In the article, McDonald describes the results of sampling 10-Ks during 1994 to 2008, and finding that almost three quarters of the words identified as negative by the Harvard Dictionary were words not considered negative in financial contexts. Some examples include: tax, capital, board, foreign, order and depreciation.

In order to combat this problem for financial sentiment analysis, the McDonalds and other colleagues sat down and went through each word in a dictionary and identified those that they felt would most frequently be seen as reflecting a particular type of sentiment or attribute. They then back-tested the lists by creating word counts based on all 10-K filings and carefully examined words that seemed misclassified, typically due to alternative uses.

Dr McDonald has also created six dictionaries behind a Dow Jones product called Lexicon, which uses the proprietary sentiment dictionary to create custom sentiment models. It is actually financial information and data providers, such as Dow Jones, that have expressed the most interest in FINIF, aside from investment and hedge fund managers.

Though FINIF had at one time operated on a subscription model, it has since pulled back in order to further build out the database and possibly partner with others in order to scale the business. So, I would say you could classify it as in post-Beta development.

FINIF certainly has taken on a niche in sentiment analysis and drawn some attention. It will be interesting to see how things unfold for the McDonalds and FINIF in the near future.

By Heidi Longaberger

Heidi Longaberger is currently Research and Information Manager for NorTech, a technology-based economic development organisation with a focus on Northeast Ohio. Heidi gets to provide innovative advanced energy and flexible electronics companies with market intelligence and analysis, while providing input on NorTech's strategic initiatives. She is also developing an information and knowledge management strategy for NorTech, and gets the opportunity to write about economic development issues for the region. Heidi's background in information and research includes extensive work in the telecommunications and venture capital industries. Heidi can be reached at Heidi@Longabergerinfo.com.