semantic search | Advocate's Studio

When one thinks of the practice of law, I imagine one thinks “qualitative”, relegating the “quantitative” to the accountants and actuaries. But the more data we can collect and mine, the more likely even the most judgment-based tasks can be influenced by trends. That is the premise behind the label “quantitative legal prediction” – mining “big data” for trends in legal decisions and court filings, statutory evolution and legal billing and task performance in order to predict outcomes. The ABA Journal reports on an article at Law Technology News touching on the concept and how it is being used in a few scenarios. How could this work? Imagine crawling the opinions and decisions in Pacer to find the arguments that supported the most winning results? While the key to the effectiveness of the tech is getting the information into mine-able form, there are already services and firms making inroads. There are programs for e-discovery using algorithms to identify documents most likely to be relevant to a given discovery request. Ty-Metrix, a legal billing software, has collected massive amounts of billing data and can now mine it for law firm rates and the factors that affect those rates. Then there is Lex Machina, an organization that has spent 10 years trying to build and organize an effective database for intellectual property litigation.

There are other examples as well in the LTN article. And the article and its examples herald for me an era in law practice that I have been eagerly anticipating – that moment when we can dip into all the available information on a given subject, tap the data and return predictive answers back using an interface like Wolfram Alpha. I don’t believe that this ability means the death of human judgment. Quite the opposite – armed with better data, more complete information, our judgment will be sharpened and improved. I welcome our Big Data Driven overlords and their semantic minions. It will be interesting to see what the next few years brings, especially if efforts to demolish the PACER pay wall come to fruition.

They say that knowledge is power, and rightly so, particularly in the Digital Information Age (my term, FWIW). Access to information is important, but being able to leverage via machines the extra step that links the pure data to contextual relevancy is the current Holy Grail of Search. Pioneers in the digital knowledge game like Wolfram Alpha and Siri have been making extraordinary inroads in pairing correct answers to natural language questions. Semantic search – the ability to parse contextual meaning from a search inquiry by making connections across data sets – is the key to the next step in the evolution of search.

So, where is Google, the de facto King of Keyword Search, in all of this? Well, as of yesterday, right in the thick of it apparently. Google has introduced a major new refinement of its venerable Search Product called Knowledge Graph. Knowledge Graph appears to be a matrix of contextual connection behind the pure search terms that assist Google in showing results that make sense, as well as direct answers to queries right on the search results page. Instant results will highlight the answer Google believes you intended to find, as well as other possible answers to your question that make sense based on context – the connections between data points. The example from Google’s blog post debuting Knowledge Graph is the phrase “Taj Mahal”, which could be a monument, a Grammy award winning singer, a casino or the Indian restaurant down the street. Before, Google’s search would simply turn to its vast store of crawled data to find sites where the words “Taj” and “Mahal” appeared near each other, putting the sites that had the most clicks for those keywords at the top of the list. With Knowledge Graph, Google takes the next logical step by “guessing” the meaning you intend when you type “Taj Mahal” and presumptively returning relevant results. Pretty freaking cool.

To make this happen, Google is leveraging content stored in trusted sites, such as Wikipedia, Freebase, the CIA World Fact Book and other locales. Not unlike Wolfram Alpha, which turns to its own internal knowledge base comprised of data from official public or private websites, and systematic primary sources.

There are three main features of the new Knowledge Graph. First, searchers will see different collections of results accessible via one click – click over instantly and tell Google which segment you are interested in researching. New summary info provides information on people, places and things right on the search page, obviating the need to click through to Wikipedia – good for quick bits of information, leaving you free to click through to get more detail if you need. Finally, Knowledge Graph takes it all one step further by providing the second tier information that users tend to look for after making their initial search. Google apparently is able to map those secondary searches and make the information easier to tap into, collapsing first and second searches down and improving search efficiencies. Google also shows other searches that people commonly made when searching for the same information. Google has accomplished the corralling of data in such a way that it can parse likely intent and direct searchers along the search path in a reliable fashion.

Is this all good? Well, not quite and definitely not for power searchers. Yet. Google’s new toy will work best with people, places and things and mostly likely with well-known people, places and things. More arcane and obscure information likely hasn’t been properly mapped yet, particularly since it appears Google’s tool depends on what lots of other searchers tend to do. Which raises an additional question regarding what lots of other searchers tend to do – if you are not your average searcher looking for not your average information, you might find the Knowledge Graph more hindrance than help at this point. However, I wholeheartedly applaud Google’s efforts (as well as Bing’s similar effort released earlier). There is definitely a place for instant, contextually-relevant results in everyone’s search plan. My sense is that it will REALLY get interesting when contextual, semantic search can delve the deeper recesses of data and make finer connections. Like the connections our billions of neurons make when we cogitate on a problem or try to recall key information. I can hear Majel Barrett’s voice now. Is the age of Artificial Intelligence upon us? Maybe. Just maybe.