How computational linguistics went from Cold War communications to transforming enterprise work

How has computational linguistics evolved and transformed for enterprises today?

digital transformation
Thinkstock

The Cold War is often cited as the starting point for the scientific work around natural language processing (NLP) that has since evolved into computational linguistics, an artificial intelligence (AI) technology now in commercial use. The story goes, the phrase “The spirit is willing, but the flesh is weak” once used to test the work that had been done, translated to Russian and back to English as “The vodka is strong, but the meat is rotten.” To say this technology’s earliest forms had challenges is quite an understatement, but the reality is we have come a long way since then. Businesses today inform risk, operations and performance decisions in the same way we had, with limited success back then, aspired to inform diplomatic, political and military actions. 

Language, context and reasoning via computational linguistics

Over the last several decades, language processing science has evolved in two complementary areas. First, advanced NLP today is shaped by deeper learning algorithms relying on computational statistics that enable a machine to understand content using mathematical relationships from past word associations. Second, computational linguistics, which enables the machine to understand content by deciphering its linguistic structure, has emerged as a useful enterprise AI technology that extends the applicability of language processing.

Because computational linguistics can work from small amounts of data, it is particularly effective in solving specific and narrow business problems that span operating processes. When we read something, we understand the content contextually – either discovered in the document (e.g., how it is used in the sentence) or with a priori knowledge (e.g., is it a press release or legal contract). Discovering this context requires a multi-pronged approach for a machine, but when done right, it dramatically enhances understanding of the content. And it does not require a hundred thousand documents to understand the next one.

Another advantage of computational linguistics is that it provides traceability to the source of the decision that the machine made. This “non-black box approach” is essential for mission-critical applications, particularly for companies operating in regulated markets. 

Enterprise benefits: practical examples

While perhaps not as exciting as international espionage, computational linguistics nevertheless offers businesses competitive differentiation and performance advantages that completely transform how enterprises interact with their customers, drive top-line growth, and reduce operating expenses. Let’s look at a few examples:

1. Risk management in commercial lending

Risk management in commercial lending traditionally involved reading through each of the balance sheets and associated financial statements of the assets in the portfolio and calculating risk scores that are then aggregated across the portfolio. Today, forward-thinking companies are using computational linguistics to extract data from thousands of balance sheets – often in multiple languages and across different accounting standards – and dynamically calculating risk across the portfolio. This approach not only improves accuracy and efficiency, it makes risk assessment more responsive and more comprehensive, completely transforming the underlying business value drivers.

2. Contract reconciliation and enterprise performance

Most enterprises end up with thousands of contracts and multiple invoices associated with each contract. Computational linguistics’ ability to read all the contracts by machine, and extract the relevant parameters (e.g., shipping rate) and associated values (e.g., $30/express delivery) converts all that unstructured data into a usable, structured format. This is also done similarly for all the invoices corresponding to those contracts. Once all this information is available in a structured data format, it is easy to run simple reconciliation algorithms that can identify over-billing and drive better economic performance.

3. Wealth management

For institutional investment firms that deal with complex custodial statements each with hundreds of transactions across a broad set of instruments, including hedge funds, derivatives, and specialized funds, calculating and delivering performance reports is traditionally a cumbersome and difficult process. Machine reading these documents with computational linguistics provides the ability to extract all the relevant asset and trade information dynamically, and easily convert this data into automated portfolio performance reports. This cuts down on expense and effort. As an example, this process used to take up to 90 days; with the use of computational linguistics, forward-thinking wealth management companies process all this overnight. Furthermore, these reports have become configurable and easily customizable, delivering increased business value to the institution’s customers, and transforming its overall service delivery.

Digital transformation with artificial intelligence

We have come a long way with artificial intelligence technologies – from conversational AI with our Amazon Alexas to computer vision on our iPhone photo libraries – we see it and feel it in our personal lives. What we do not as readily see is how so many traditional business processes are being completely reimagined on the back of data analytics and artificial intelligence technologies.

But the reality is entire business value chains are surely and certainly transforming themselves and as more and more corporate boards now own and drive the digital agenda, it is heartening to note we have progressed substantially since the Cold War’s early days of challenging translations.

This article is published as part of the IDG Contributor Network. Want to Join?

NEW! Download the Fall 2018 digital issue of CIO