For anyone who’s read the news recently, it’s hard to avoid the topic of artificial intelligence (AI). In recent months, the world has been stunned by both achievements and gaffes in the field: from AlphaGo’s victory over one of the best human players at Go to Microsoft’s notoriously racist online chatbot Tay, AI has been firmly on the media agenda.
However, to most businesses, these projects are simply not relevant. These applications are divorced from real business needs, which is frustrating, given that the underlying technologies of natural language processing and linked data technology have such great potential.
News just in: big data? Still too big
The phrase big data has been around for a long time now, but it is still appropriate for describing the problem facing businesses today: companies find themselves sitting on an ever-growing pile of digital information, and using this data effectively is a significant IT challenge.
Yet, if we look at today’s major technology players’ advancements, the eye-catching research isn’t exactly aimed at solving this problem. For example, Google recently published details of its experiment analysing literature and authorship styles across the centuries. This is clearly a first step towards something very interesting. However, it’s light years ahead of what most businesses’ IT is capable of implementing in order to boost their bottom line.
A business with the ability to do relatively basic natural language processing with its data would already have a competitive advantage in most industries, never mind making use of Google’s cutting-edge research described above. This is particularly relevant for companies across the publishing, pharmaceutical, legal and financial services industries - for these companies, the ability to identify entities and relationships within written content in an automated manner will deliver massive operational benefits. From searching for patents, extracting research from papers or assessing financial products for reconciliation and compliance, companies performing these actions are held back from meeting their potential by the burden of leveraging data effectively.
What is natural language processing?
Put simply, natural language processing technology allows businesses to analyse their text data in a more nuanced way by identifying entities within text and the relationships that link them together. This is then saved in a particular kind of database as linked data: one of the key business benefits of linked data is that it empowers human users to glean insights from the data that would otherwise either not be achievable within the same timeframe.
This technology is a long way behind headline-grabbing AI projects such as Google’s literature analysis project or, for example, Microsoft’s recent Tay AI disaster, but that doesn’t mean it is unimportant. These AI projects are driven by technology which enables natural language understanding (a complex expansion of natural language processing), and this technology is simply not developed enough to be useful for IT teams in businesses today.
The sheer simplicity of the actual experiment Google’s natural language understanding research group conducted demonstrates the technology’s lack of maturity. Microsoft’s Tay was transformed from innocent teenager to aggressive racist within 24 hours, showing that while the AI bot was able to process the language the trolls were feeding it, it was not able adequately to understand the content of that language or the social and political contexts that language exists within. Clearly, there is a lot of work still to be done.
By contrast, natural language processing, linked data and semantic technology are already mature technologies that are being used by businesses and corporations around the world. Not only is this technology already in use, but there are already established standard languages, such as RDF, to ensure interoperability.
In-house IT teams are often aware that they have a problem with their data processing and flexibility, but with so many news stories about the data processing possibilities AI technologies can bring, it can be difficult to assess what solutions and architectures to sign up for. The best method to tackle business data processing problems is to focus on the technology that works today, not what may work tomorrow.