[This article first appeared in the Clarabridge Bridgepoints newsletter, Q1 2008.]
Text analytics has built on early successes in fields such as the life sciences and intelligence to win acceptance in a broad variety of industries. That the financial sector is among the most significant of these is no surprise. “This industry relies heavily upon FinTech vendors to support their businesses,” according to Jeanne Capachin, IDC research vice president, Global Banking. It represents over 20% of global information technology (IT) spending according to a December 2006 IDC research report.
The financial sector includes banking, credit, financial information, insurance, investing, mergers & acquisitions/due diligence, payments, and trading markets. It serves retail (consumer) and institutional (corporate) markets and includes informational, service, and regulatory components. These markets and sector components produce and consume very large volumes of structured transactional data, “semi-structured” forms and filings, and “unstructured” news and communications.
Business intelligence, data mining, and predictive analytics have long been important contributors to the financial industry’s bottom line, delivering significant competitive advantage to forward-looking firms. The sector relies on these analytical technologies, which have traditionally exploited only fielded, numerical data, to identify opportunities, manage risk, control costs, and better serve customers. Given the possibility of adding “unstructured” and “semi-structured” sources to the mix, the financial sector has been quick to adopt text technologies.
Text analytics supports both “horizontal” enterprise functions such as customer relationship management (CRM) and marketing and “vertical” needs that are specific to particular business domains, tailored to their goals, information sources, and workflows. Financial-services firms can benefit from both applications of text analytics, that is, from both Voice of the Customer and Enterprise Feedback Management solutions that analyze the diversity of customer communications, where capabilities are not industry-specific, and solutions that mine financial reports, analyst reports, regulatory filings, news articles, and the like specifically for financial information about corporations, individuals, and market conditions.
We will look in more depth at the application of text-analytics to financial-information sources in support of better financial decision making: at ways solutions support lending, investing, insurance, marketing, and trading activities, and at the provision of financial information for consumers and businesses.
Mining financial-information sources
Typical text-analytics processes start with information retrieval and extraction – by identifying promising sources (where semantically enriched search can help) and discerning significant entities, relationships, and sentiments as well as attributes that describe these items, which are sometimes collectively known as “features.” Entities of interest include the names of companies and individuals, trading symbols (e.g., GOOG for Google), locations, dates, and monetary values. Discernable relationships could include anything from membership of an individual in a corporate board of directors to revenue statistics (“Q4 sales were $247 million, representing 7.4% year-on-year growth”) to a price quote for a security (“Bid: 60.15 x 1000”). Sophisticated linguistic, statistical, and machine learning techniques boost the accuracy – the precision and recall, measures of relevance and completeness – of text-mining steps.
Text technologies also support automated classification, summarization, and routing of documents; tasks that complement information extraction. They extend the financial analyst’s (or consumer’s) reach, speed processing, and lower costs in both the extraction and automated document handling scenarios.
Financial analysis is not a new practice; however, so it is important to integrate these new technologies with established software tools and work practices. One approach is to support extraction, not just for “workbench” style analyses of information from textual sources, but also for integrated analyses of text and data. Extraction of text-derived information to a relational database would allow integrated analytics using the reporting, OLAP, data mining, and visualization capabilities of familiar business intelligence and advanced analytics tools. A focus on databases opens up additional analytical possibilities. These include the computation of new performance indicators that consolidate operational measures derived from transactional systems with data extracted from textual sources, the creation of predictive creditworthiness and risk models that are scored based on data extracted in real time, and financial forecasting from text-derived data. The end goal is creation of actionable financial information.
Programmatic (algorithmic) trading provides an interesting alternative use case, one that takes text analytics to the leading edge of real-time decision making. Automated trading relies on market-data feeds, which has given rise to a new software category of data- and event-stream processing tools. These tools crunch high-volume data feeds with very low latency, joining information acquired in real-time to historic data stored in databases. Traders ignore news feeds at their peril because breaking “real-world” events as varied as a terrorist attack and the news of a corporate merger can cause pricing anomalies, that is, opportunity. Look for growing adoption of real-time text analytics as the technologies continues to mature, driven by financial-sector demand.
The financial information market
A last text-analytics application to consider is in delivering value to the $35.5 billion financial-information market. According to research firm Outsell, Inc., this market accounted for 9.8% of the total 2006 information industry, indicating its importance. The financial segment covers the provision of consumer and company credit information as well as financial information services purchased by clients such as banks, insurance companies, and investors. Market leaders include Thomson Financial, Bloomberg, and Dow Jones/Factiva; all have strong text-analytics initiatives underway for content generation, content provision, and advanced search and information retrieval. The use of text technologies at these organizations is an indicator of the value text analytics can bring, not only to their client, but to the broad market of financial-sector organizations seeking to extract maximum value from financial information.
One thought on “Financial Sector Text Analytics”
Please describe the significance and application of text analytics in finance in simple english.