About two years ago, I read a compelling book by Malcolm Gladwell called The Tipping Point: How Little Things Make a Big Difference (Back Bay, 2000). Gladwell puts forward the social networking concept of how little ideas become huge tidal waves. He describes several key personality types, including connectors, mavens and salesmen.
Connectors are the socialites who know a lot of people and automatically link them together to rapidly spread ideas through many communities. People just naturally want to know and interact with connectors. Mavens are the information gatherers of the social network. They gather ideas and spread them through their communities of interest. People just naturally want to hear what mavens have to say. Salesmen are the
persuaders who sell those ideas to larger and larger groups until the little ideas tip to huge tidal waves. People just naturally want to buy from salesmen. Figure 1 (page 9, KMWorld, Vol 16, #5) shows the social network of mavens, connectors and salesmen.
The idea of the human connector resonates because it highlights some key issues for knowledge management, particularly in the areas of relational analysis (about the relationships between entities), also known as social network analysis and link analysis. Human connectors seem to embody those concepts; connectors are able to automatically identify, manage and search the social network and relationship information they have within their control. They seem to do that as part of their normal existence, and it is as natural to them as eating and breathing. However, in computer science and knowledge management, the issue is how to gather the right metadata about the entities and relationships within the
information and data being processed.
In fact, the computer science discipline of relational analysis focuses on the same concepts as those brought forward in The Tipping Point. As a technique in knowledge management, relational analysis helps us in mapping and measuring relationships and information flows among people, groups and organizations.
Relational analysis is a method for defining networks of people or entities and their relationships. It helps us understand networks of people, who the participants in the networks are, and how the participants or actors are positioned in their respective networks. Measuring networks, finding the centers of networks and their end-points, gives us insight into the important individual roles, organizations, locations and groupings in a network. Through the use of metadata, social network analysis defines how connectors, mavens and salesmen, among others, fit together in their networks. It shows how they relate to one another, who has the most and least influence, how information clusters and more.
Automatically generating, processing and visualizing that kind of business intelligence information is crucial to improving business decision-making and to boosting productivity. When companies can quickly identify the right connector or the right maven or the right salesman, they make the right decisions the first time. That means products are developed faster, goods are produced more efficiently and customers are more quickly serviced (and hopefully more quickly satisfied).
For the past 6 years, I have been working with new technologies that can automatically generate relational metadata from text documents to identify relationships between people, organizations and places, and what those relationships actually are about. Over that period, computer science techniques in metadata generation have come a long way, bolstered by improvements in hardware processing power and memory, and in software development in artificial intelligence, natural language processing and telecommunications—particularly the World Wide Web.
Trustworthy software systems—accurate in their depictions of people, organizations and places, and their relationships—are the result. I’ve had the good fortune to be exposed to a variety of different products that provide a reliable level of automation in terms of generating relational metadata, allowing people to manage, search and visualize that metadata.
Relational metadata and search
A number of companies provide information extraction tools that extract a wide variety of metadata from unstructured documents such as MS-Word, MS-Power Point, PDF, Web pages, etc. Those tools include capabilities such as categorization, clustering, summarization, geotagging, entity extraction, transliteration and, of course, relational metadata about people, organizations and places.
Attensity is a leading company offering technology to extract relational information from text documents automatically, which requires advanced analytical capabilities that use natural language techniques, based on artificial intelligence and heuristics.
Using natural language approaches, Attensity provides a parsing engine that understands eighth-grade grammar. It parses text in order to extract the nouns, verbs, adjectives, adverbs, prepositions and other parts of speech that describe all of the relationships within a document or set of documents. Attensity generates a principal form of relationship metadata called an actor action object (AAO) triple, which describes the people, places and things in a document and their relationships.
Attensity has been used commercially in a variety of different business applications. One of the most compelling uses has been to find the real problems manufacturers face with warranty claims. Attensity found that in most call centers, customer service representatives have a call management system that uses keywords to categorize a customer’s specific problem. Due to time constraints and productivity requirements, customer service reps often pick the best fitting general problem category and then record the essence of the problem in the comments area of the form. Most companies don’t have the tools to gain access and insight into that important repository of information, so that gold mine, which directly impacts customer satisfaction, goes undiscovered.
While manufacturers have found that 90 percent to 95 percent of all calls are relatively well categorized using the standard categories, about 5 percent to 10 percent of the problems (including serious ones) end up buried in the comment section. Using Attensity’s ability to pull out structure and relationships from the unstructured, text-based comment section of the call center records, companies are able to determine what the real problems are and more rapidly respond to customer needs. The solution often discovers small pockets of significant, hidden and costly problems. Identifying that kind of information boosts customer satisfaction and helps productivity on the manufacturing side because hidden problems are found more quickly.
Attensity’s ability to parse text and pull out links and relationships can be displayed using a variety of visualization tools. For example, Figure 2 (page 9, KMWorld, Vol 16, #5) shows i2 Inc.’s Analyst Notebook iXv Web plug-in, visualizing AAO link information produced by Attensity’s Link Factory. Link Factory is designed to help analysts perform relational analysis on entities and entity relationships.