Hal Marcus

Hal Marcus
Hal is the Director of Product Marketing for OpenText Discovery. A licensed attorney, Hal practiced as a Wall Street litigator before commencing a career in technology that now spans over 20 years. He writes about practical applications of artificial intelligence, machine learning, advanced analytics, and business intelligence, particularly in the realms of information governance, e-discovery, and compliance. (Anyone with the name Hal is bound to have an interest in AI.)

When Search Meets AI, It Takes You Further

AI

Many think of AI as human-like robots or human-sounding voices that interact with actual humans, but the real value of artificial intelligence may derive from what it’s doing behind the scenes. For example, when unsupervised machine learning (a species of AI) is applied to enterprise content, it can develop a deeper understanding of that content and deliver insights to humans in instantly valuable ways. Content is about more than mere words; it’s about how words interrelate. By analyzing the statistical co-occurrence of terms across enterprise content from a range of sources, machine learning forms sophisticated models that can take search much further. It can identify concepts, extract phrases, suggest better queries, and pinpoint internal SMEs with relevant expertise—whether or not their written profiles even reflect it. AI-infused search helps people find what they’re looking for, even when they’re not exactly sure what that is. This is what OpenText™ Decisiv is all about. “Asia Not Asia” Imagine that you’re looking to understand more about your business operations in Asia. Your first instinct might be to search your various data stores for documents containing “Asia,” but will that term necessarily appear in the most valuable content? For that matter, how many documents will list city, state, country, and/or continent? Without machine learning, you would need an exhaustive search taxonomy to address all these variations. When a Decisiv user types “asia,” the system instantly and automatically retrieves documents that (1) contain the word Asia and/or (2) are conceptually related to Asia. This not only casts a wider net, it provides results based on a more sophisticated relevancy analysis than simple keyword searching could hope to deliver. Typing in the seemingly illogical query “asia not asia” is illustrative; it shows the user content that is conceptually related to Asia, but doesn’t contain actually contain the term Asia—displaying only the documents found by unsupervised machine learning above and beyond keyword search. Concept grouping categorizes documents according to linguistic patterns that we humans have a hard time identifying across large volumes of data. Machine learning makes such conceptual analysis automatic and highly scalable, and humans get the benefit. Further Still Concept groups can also propel a researcher to new areas of useful content. By looking at a list of the key concept groups that appear in response to a search or metadata filter (along with the document counts for those groups) researchers can see an overview of the content that’s available, spot pertinent aliases, adjust their search terms, and include or exclude concepts to meet their objectives. Concept groups are displayed with characteristic labels (top words and phrases that appear in those groups) to easily provide a sense of what each group represents, and how useful it’s likely to be for a given search. Take Decisiv Action Register for OpenText’s annual user conference, Enterprise World, in Toronto this July to learn how AI-enhanced search can help empower your digital transformation. You’ll hear from top users of Decisiv Search and our other award-winning Discovery products how they’re leveraging machine learning for more effective enterprise information management. All humans welcome.

Read More

Why Lawyers are Adopting AI Faster Than You

AI

When you think of bold, innovative users of transformational technology like artificial intelligence (AI), you naturally think: lawyers. It’s obvious, right? Risk-averse, measured, charge-by-the-hour, brick-and-mortar professionals that parse the written word (“Heretofore? No, hitherto!”) and deliver cautious, nuanced advice (“I didn’t hear your question, but regardless my answer is: It depends.”). Who better to make practical use of today’s cutting-edge AI? (“Alexa, draft an amicus brief in support of my motion in limine. Please.”) Lest the irony be missed, the legal industry is deservedly notorious for being a technological step or two—or more—behind its clients. Yet law firms and savvy corporate legal teams have been pioneering the use of artificial intelligence since the last decade. There is not a litigator of note today that hasn’t heard of Predictive Coding or Technology Assisted Review. These terms refer to the use of machine learning to mimic an attorney’s decision-making in the context of legal discovery, the process of identifying and reviewing up to millions of documents to determine which must be produced to the other side in litigation or an investigation. Predictive Coding can mean dramatically faster and more accurate document analysis and review. Why are lawyers leveraging AI for document review? Big Data: The growing amounts and kinds of data generated by workers—in office programs, cloud apps, chat systems, shared workspaces—means an ever-increasing challenge for legal and compliance officers. To them, all of this work product is potential evidence. Bigger cost: Of the more than $200B spent on litigation across the US annually, 70% is spent on discovery, and 70% of that discovery spend goes to document review. So, anything that can accelerate or reduce review means substantial savings for corporate clients. Irrelevant content: No one likes reviewing irrelevant data. (Imagine if you had to carefully read your junk email before deleting it.) Front-loading relevant content makes document review more engaging for attorneys, which improves their productivity and accuracy. The need for speed—to insight: Over 95% of civil cases settle, as the uncertainty and cost of a trial is generally to be avoided at nearly all costs. Finding the evidence that proves or disproves your liability early on is key to negotiating a favorable settlement. Think Netflix or Pandora on steroids. Predictive Coding is about finding more like this, where this is a piece of unstructured data (an e-mail, slide deck, letter, memo, etc.) and the more like are documents that are conceptually similar—even though they may not contain the same words that made this relevant in the first place. Documents that are similar in concept but use substantially different language can be equally significant for litigation and investigations. That’s why Predictive Coding goes far beyond traditional Boolean keyword search. To enable Predictive Coding, the system performs statistical analysis on the co-occurrences of all the words in each document ingested, even across millions of documents. It then creates sophisticated models around a handful of documents judged by attorneys to be relevant to the issue under review. It looks across the data set and finds more documents closely related to those models and suggests them to the attorneys for priority review. As attorneys review the suggested documents and label them relevant or irrelevant, the system gets smarter, refining the document models for even better results in the next round. With Predictive Coding, attorneys can find virtually all the relevant content in a data set by reviewing just 10-30% of it, shaving off weeks or months of tedious review and surfacing critical evidence far faster. What OpenText is doing about it: In 2016, OpenText acquired Recommind, a pioneer in advanced analytics for the legal industry for over 15 years. With unparalleled Predictive Coding and other unique capabilities, OpenText™ Discovery Suite helps enterprises discover what matters in their data—faster and more accurately. 2017 is poised to be a banner year for legal technology, as awareness and experience with Predictive Coding are approaching critical mass. Our vision is to see machine learning used to add value to every matter, on virtually every data set. After all, who better to drive technological innovation than your venerable counsel?

Read More