Text analytics: not just for customer sentiment
Analytics plus—proficiency in language
In general, text analytics solutions can handle multiple languages, but usually the software producer has created an interface for each language and the product is analyzing documents in that language. Babel Street has a new approach with Babel X, which can search in English or across 200+ languages and translate on the fly. “Babel X can identify each language and perform entity extraction,” says CEO Jeff Chapman. “For example, it can find ‘Coca Cola’ in any language even though the words for it are sometimes different in that other languages.”
Geared for the “volume, variety and velocity” of big data, Babel X can collect a thousand documents per second, show (for example) all the ones in Chinese that have a negative sentiment about a food product in near-real time, and build a word cloud that shows the topics that are co-located. That analytical process indicates whether the food problem is bad taste, packaging or other problem and identify the root cause quickly.
Designed to run in a highly unsupervised mode, Babel X does not need a training corpus. “If ambiguity is found,” notes Chapman, “then human intervention can take place and a rule can be written to resolve the ambiguity.” Babel X includes a multilingual ontology of more than 900 million unique entities that supports concept searching and provides geo-inferencing. It is well suited to analyze streaming internet traffic because of its speed.
After the Boston Marathon bombing, Babel X was employed to search for relevant social media messages, using geolocation to focus on messages sent near the site of the bombing. Babel X then looked retrospectively at the preceding 24 hours and analyzed social media messages for references the bombers made to other individuals and topics, which led to the development of a profile.
“Babel X does not require a data scientist in order to produce results,” Chapman says. “Users can easily screen for names, dates, times, keywords, number formats and even emojis.” It can search across more than 40 social media platforms, more than 1 billion blogs and message boards, and any designated URL.
“The system is looking for things in context. For example, if there are reports of a chemical in food that might be related to cancer, it will look for cancer,”
Besides being able to analyze text, the solution needed to be capable of predictive analytics and ?enable visualization of the data so that someone not familiar with the data analysis could understand the results.