Is an ‘Insight Engine’ a Bunch of Hooey? You Decide
This is the conclusion of the two-part blog, Is an Insight Engine a Bunch of Hooey?
Since all enterprise search vendors are now ‘insight engine’ vendors, one has to be careful as Gartner has opened up the ability to make noticeable claims about a search engine, just by renaming it to an insight engine.
For example, some ‘insight engines’ gather statistics based on how often a document is accessed, and place it higher in the search results. This is, at least to me, is faulty logic. 55 percent of end users will select the wrong document more than once. I do it myself, sad to say. Over time, the information retrieval process could just be cluttered by unwise end users, like me, who keep clicking on the wrong document – again, and again, and again. Insight or hooey?
Some of the vendors provide ‘pre-packaged’ taxonomy structures for specific industries, but lack classification rules or the ability to easily add or modify rules. Some of these vendors are now touting artificial intelligence (AI) as an offering. They weren’t a year ago. Anyway, these applications often do not have the flexibility to be customized to the unique nomenclature of an organization. This is actually a feature that is easily overlooked, yet it is critical to accurately retrieve content in the language that it is spoken or written within an organization. Insight or hooey?
Machine learning, or artificial intelligence, can teach itself to change or grow as new data is added. The downfall is that instead of extracting data for human comprehension, as is the case in data mining applications, machine learning uses that data to improve the program’s own understanding, rather than the human understanding.
The problem I have with search vendors throwing around these buzzwords is that most of them are fibbing. Seriously. Also, machine learning is really good at entity extraction. In fact, our software uses natural language processing (NLP) for just that purpose, but it is not the core technology. In a way, I think of it as taking the content outside the realm of ‘unstructured’ and putting it in ‘structured,’ because it fits neatly into pre-defined fields. I don’t think that’s the best way for a search engine to behave – every piece of content is not an entity. Insight or hooey?
Semantic networks are expressed visually, but they are intractable for large domains, and do not represent performance or meta-knowledge very well. Some properties are not easily expressed using a semantic network, for example, negation, disjunction, and general non-taxonomic knowledge. Expressing these relationships requires workarounds, such as having complementary predicates and using specialized procedures to check for them, but this can be regarded as less elegant. Insight or hooey?
Other products require organizational staff to learn a new programming language, or depend on external consultants, and mean ongoing costs if taxonomies are to be kept up to date. Some work only with the SharePoint Term Store – I would raise a red flag on that one. Some boast the ability to use Boolean expressions, proximity, or fielded search, and call that AI. Where does that come from? Some require iterative testing for every term, until you are ready to pull your hair out. Unless you can gather a highly accurate pool of training documents, you are kind of out of luck. Insight or hooey?
And my last one. I am seeing insight engine vendors starting to claim their offerings are capable of big data, or as Gartner calls it big content, analysis. The term has been employed to the point of uselessness, as have NLP and AI. Ask for a demo of that one folks!
What’s the verdict? Insight or hooey? If you want to read about a real insight engine, then click here.
Our webinars also address the topics explored in our blogs. Access all our webinar recordings and presentation slides at any time, from our website, in the Recorded Webinars area, via the Resources tab.