chris - Fotolia
Machine learning technology is making strides and offers potential for companies to deliver increasingly tailored, user-specific online experiences.
Machine learning seeks to discover meaningful patterns within large amounts of data. Coupled with continuing innovations in cloud computing and big data analytics, this technology could radically change how enterprises manage digital content to produce innovative online experiences.
An important shift in outcomes and intentions is underway. Whereas the traditional Web is largely a straightforward publishing environment, online digital experiences increasingly incorporate a variety of channels for richer, more personalized interactions. But providing these experiences requires a sound information architecture and an understanding about how relevant things are interconnected.
Beyond the Google Knowledge Graph
Consider how Google generates answers to questions. For the past several years, Google has relied on its Knowledge Graph, a large-scale effort to collect information about objects in the digital world and map relationships among things.
Ask for information about "Leonardo da Vinci" and Google displays a panel next to the search results list summarizing his basic biographical information. Included are thumbnail links to some of his most significant pictures as well as several important artists that other people have searched for.
Behind the scenes is a semantic infrastructure. Google entices website owners to contribute to its Knowledge Graph by defining tag sets for search engine optimization (SEO). Thus, many art museum curators now mark up their digital collections with standardized tags, providing the artist's name and other attributes. When they make this content accessible to Google, they automatically add the metadata describing their digital assets (such as their da Vinci images) to the Knowledge Graph.
Now come machine-learning capabilities. Google is increasingly able to answer questions that require some reasoning, such as, "Who was president when the Angels won the World Series?" Using pattern recognition technologies, Google breaks down the query to compute the semantics of each phrase and discern the intent. It then traverses the Knowledge Graph to find the right facts and suggest a relevant answer.
Google is also adapting its machine learning capabilities to other applications in its portfolio. With Smart Reply, Gmail can automatically generate reply messages, based on content and a receiver's likely response. Google Photo includes visual searching, identifying one thing in an image and matching it to comparable things in a photo collection, without manual tagging.
A virtuous circle
Google demonstrates that there is a virtuous circle underlying machine learning innovations. Predefined things, coupled with powerful pattern recognition capabilities, can produce ever more useful online experiences.
Content enrichment, through standardized tagging, adds an intelligent foundation to an information stream. The more patterns within a collection, the more a smart machine can recognize them -- bringing new experiences for end users, as well as opportunities for developers.
Capturing developers' attention
From a technology perspective, the race for enterprise developers' attention heated up in 2015. Google, Facebook, IBM, Apple, Microsoft and Amazon have all made at least some of their machine learning algorithms available to external developers, often as open source tool kits.
In addition, IBM continues to release cognitive computing services that incorporate various pattern recognition capabilities. Application developers can now call Watson Developer Cloud Services to analyze text, extract concepts, classify natural language, identify personality traits, transform speech to text and text to speech, find things within images, and perform a host of other thought-like functions.
Building local knowledge graphs
But machine learning within an enterprise requires more than clever algorithms and sophisticated programming. While Google has invested resources toward semantic infrastructure to meet consumers' expectations on the public Web, building a smart machine for an enterprise remains a challenge.
Machine learning is based on prior knowledge. Beyond the algorithms are innovative ways to tag content and distribute results across like-minded organizations. One answer is enterprise knowledge graphs that describe the things companies value and map connections between relevant items.
Three steps for getting started
Knowledge graphing begins by naming things in meaningful ways. Hardly a revolutionary idea, this activity takes on a new sense of purpose for next-generation online experiences, ones that can automatically detect patterns and propose seemingly intelligent solutions. Here are three steps for getting started:
1. Build business taxonomies. Start with things that drive value. Build useful taxonomies related to business functions and outcomes. For example, manufacturers have product lists, publishers have topics, and government agencies have missions and programs.
Many enterprises already define one or two taxonomies when tagging content for a Web content management (WCM) system. Initially, these are bare-bones efforts. Expand on them and catalog what's important to the enterprise. Introduce new taxonomies that capture critical business factors, such as processes, strategies and the competitive environment.
As taxonomies expand, organizations may need to add a taxonomy management tool to maintain term sets in a systematic manner.
2. Capture experience. Pay attention to the overall information architecture. Consider how key stakeholders -- including customers, subject matter experts and employees -- categorize relevant items of interest. Identify missing categories. Then either add the terms to existing taxonomies or define new ones.
Designing from the bottom-up can augment top-down definitions: Customer experiences should balance subject matter expertise. Incorporate ad hoc terms generated through free text keywords into enterprise taxonomies.
3. Map interrelationships. Identify how categories in different taxonomies are interrelated. Map the relationships. These connections become elements for an enterprise knowledge graph.
Hardly a onetime design effort, knowledge graphing within an enterprise is an iterative process, built by capturing insights and experiences over time.
Machine learning pays off through the virtuous circle for content enrichment. But organizations need to make the up-front investments. Getting your information house in order ‑‑ enriching content, maintaining taxonomies and understanding how things are interrelated -- is a good place to start.
The growing importance of Web personalization
Web upgrades open doors for customer experience strategy
Location adds context for Web personalization