May 9, 2016
Over the past several years, Forrester's research has written extensively about the age of the customer. Forrester believes that only the enterprises that are obsessed with winning, serving, and retaining customers will thrive in this highly competitive, customer-centric economy. But in order to get a full view of customer behavior, sentiment, emotion, and intentions, Information Management professionals must help enterprises leverage all the data at their disposal, not just structured, but also unstructured. Alas, that's still an elusive goal, as most enterprises leverage only 40% of structured data and 31% of unstructured data for business and customer insights and decision-making.
So what do you need to do to start enriching your customer insights with unstructured data ? First, get your yext analysis terminology straight. For Information Management pros, the process of text mining and text analytics should not be a black box, where unstructured text goes in and structured information comes out. But today, there is a lot of market confusion on the terminology and process of text analytics. The market, both vendors and users, often uses the terms text mining and text analytics interchangeably; Forrester makes a distinction and recommends that Information Management pros working on text mining/text analytics initiatives adopt the following terminology:
- Text mining extracts structures from unstructured text. Text mining is a complex process where there is a specific sequence of steps to follow. Text mining first involves connecting to data sources, ingesting the text, cleaning it, preprocessing it, and mining it to extract structures like entities, concepts, and sentiment scores. Most often the process requires several iterations for data enrichment, AKA "training." Business domain subject matter experts and professional linguists can reiteratively train the system to be more accurate with each iteration by letting the system leverage industry or business domain specific ontologies, taxonomies, and lexicons.
- Text analytics analyzes the findings of the text mining process. Text analytics then answers the question "What have we found?" It's a process that, via a graphical user interface (GUI), lets the user analyze and organize the findings. Text analytics features include displaying counts of the structures uncovered in the text mining process, their relationships (via network or graph diagrams), and hierarchies (industry, product hierarchies).
- Post-processing text analysis uncovers patterns. Post-processing text analysis analyzes patterns based on various attributes. This step of the process is used to explore time (as in "Is the sentiment trending up or down over time?"), region, customer segment (as in "How does the sentiment vary by customer segment?") and other patterns by any of the available attributes. In the world of structured data analysis, this is often referred to as online analytical processing (OLAP). In the world of search, this is often referred to as faceted navigation.
To address all of the text mining and text analytics requirements, vendors compete in a broad, diverse, and complex landscape that includes more than 200 potential players. But before considering investing in a specialized text mining and text analytics product, Forrester recommends that Information Management pros investigate whether some of their existing enterprise software platforms, tools, and applications (extract, transform, load [ETL], database management systems [DBMS], voice of a customer [VOC] and others) already have all, most, or some of the required text mining and/or text analytics features.
Next when you are indeed ready to consider, evaluate and deploy an enterprise grade Text Analytics plaform, check out our latest assessment The Forrester Wave™: Big Data Text Analytics Platforms, Q2 2016. The 10 vendors Forrester included in this assessment were Attivio, Cambridge Semantics, Clarabridge, Digital Reasoning, Expert System, HPE, IBM, Linguamatics, OpenText, and SAS.