Two Benefits of Machine Learning in the Attivio Platform

Some days, it seems that new machine learning applications are popping up everywhere you look in the news. In this author’s opinion, the search market seems to have anointed machine learning as the new hotness. This is a fascinating realization, because anyone who has spent more than a couple years deploying search in the enterprise knows that machine learning has been used and applied in exciting and unique ways for years and years.

In my experience, clients tend to conflate what machine learning means when it comes to Enterprise Search. Of course, it’s not their fault – machine learning is everywhere you look! But when Attivio says machine learning, we mean two things:

  1. Text analytics via machine learning
  2. Search relevancy via machine learning

Performing these two (admittedly difficult!) tasks ultimately enables the end user to find an answer to their enterprise-specific question.

Machine Learning in the Attivio Platform

Text Analytics via Machine Learning

First, let’s take how Attivio achieves text analytics via machine learning.

  • Directed/Supervised approach - Attivio takes a directed/supervised approach of machine learning to achieve text analytics techniques such as classification and sentiment analysis. This means that the platform will leverage an enterprise’s existing data sets to train a classification or sentiment analysis model via machine learning-based algorithms.
  • Discovery/Unsupervised approach - Attivio also uses a discovery/unsupervised approach to machine learning when performing activities such as statistical entity extraction. This means that we can automatically identify document-level entities across the corpus of an Enterprise without any training data sets or models.

Search Relevancy via Machine Learning

Attivio also leverages directed/supervised machine learning techniques to automatically tune relevancy for the enterprise. Similar to our capabilities around creating a model for classification, Attivio can create a model for relevancy. The inputs – or features – to train these models can really be anything, but most typically come from three classes of data:

  1. Document Metadata – Can consist of preexisting metadata or Attivio-generated metadata, with the goal of patterns and relations across an enterprise’s content having meaningful impact on result relevancy.
  2. Query Metadata – We extract information from the query to understand what terms and phrases were used and identify their proximity.
  3. Social Interactions – Any end-user interaction with an answer to a question can be leveraged: click-tracking, bookmarking, rating, commenting, sharing, downloading, etc. We also look at the user’s organization unit to be able to support different relevancy models for different groups in the organization, so we can provide a search experience that is more personal.

Your enterprise is constantly growing and changing. New employees, new departments, new initiatives, new content – our clients demand a platform which can keep up with their ever-changing landscape. Attivio leverages machine learning techniques in order to understand your employee’s content, understand your employee’s behaviors, and ultimately pinpoint the correct answer to your employee’s questions.

To find out more about machine learning in the Attivio Platform and see a demo, view the on-demand webinar How Machine Learning Drives Context for Cognitive Search. You can also sign up for a hosted trial.


Tweet This

Text analytics and search relevancy via #MachineLearning #CognitiveSearch