Customers talk to you through different external and internal channels. Use emStream, eMudhra's analytics engine, to aggregate conversations from these channels and mine this data to get actionable insights.
Collect real-time data and historical data from social sources like Twitter, Facebook, Facebook brand pages, etc.
Collect data from your app data stores like transaction notes, remarks, etc., or even customer profile data and analyze which group of customers is having problems using emStream.
Upload flat file data about customers, transactions, or messages, into the system quickly using Excel.
Prioritize your customer support emails and act on your customer problems before your customers go social to increase your CSAT level.
Collect expert reviews from major blogging platforms like Tumblr, Wordpress, etc. and analyze how your brand is getting reviewed by influential people.
Is your data source not listed here? Talk to us. We will help you out in getting data from your specific source.
Collect reviews from any review site; and let emStream analyze and tell you which attribute is getting a 5 star rating (positive sentiment) and which attribute is getting a 1 star rating (negative sentiment) from your customers.
The heart of emStream is our language processing capabilities which powers customer intelligence and real time engagement. This gives us ability to understand unstructured text like the way human brain does!
Sophisticated language and statistical models have been designed by a team with several decades experience in NLP research. Over several years of training on large volume data corpus, emStream has been able to derive very high accuracy levels of text mining.
By overlaying domain ontology on top of the data, emStream automatically categorizes sentiments into features of the product. emStream's NLP framework is fully proprietary.
"It is here that I have a bone to pick with John Doe, Speaking at the inauguration of Doe Foundation Hospital in New York last Saturday, he said.."
Bone - Product Type; John Doe - Individual; Doe Foundation Hospital - Facilities; New York - City; Saturday - Day
Prism's NER engine uses a tag set that comprises of Person (Individual, Group), Organizations (Government, Non Government, Public - private Organizations, Political Parties), Location (Place, District, City, State, Nation), Dates, Facilities (Hospitals, Library, Police Station), Figures (amounts, quantities etc.), Others... (Diseases in Health, Anatomical parts in Human Anatomy, Acts in Legal,Economic Indicators in Finance)
"Can we meet for coffee next Tuesday?" - Mar 7, 2015
Tuesday is extracted and mapped to date March 10th, 2015
"Very good location and party place. Miami the club inside it is super hot, even at 2 in the night the cleaner came for room cleaning. Very premium and cozy. Pasha on 24th floor is super"
The topic is identified as Hotels
Mary bought a loaf of bread. She gave it to John. He was very thankful to her. John went with the bread and gave it to his mother. She was very happy.
Mary bought a loaf of bread. She [Mary] gave it [a loaf] to John. He [John] was very thankful to her [Mary]. John went with the bread and gave it [the bread] to his [John] mother. She [his mother] was very happy.
Prism's sentiment engine co-relates sentiments to features/attributes of product, handles comparisons and distinguishes context specific sentiment as in the case above
"We had a session for top management of the company. Quite organized, good ambience and no interference. The food was tasty with good variety. Lots of variety in the buffet. Staff was ready to help. This facilitated learning"
Themes are identified as follows:
Top Management Good Ambience Good Variety
Mary bought a loaf of bread. She gave it to John. He was very thankful to her.
|Actor||Action||Argument 1||Argument 2|
|Mary||Bought||Loaf of bread|
|Mary||Gave||Loaf of bread||John|
Contextual NLP extends graph theory techniques to differentiate entities, disambiguate them and resolve them against available knowledge bases. emStream's technology has been built to resolve against public knowledge bases such as Wikipedia, WordNet, etc. and can be configured to lookup private knowledge bases.
"The president gave a brilliant speech during his trip to Iowa in June 2014"
Names of people are typically resolved against public knowledge bases contextually. In the above case, emStream's technology uses the location and date as a context to determine that this refers to President Obama.
Business Intelligence has seen a paradigm shift where mining insights are not nearly enough but how do use these insights to arrive at predictions about customer behavior, potential purchases, other outcomes that could directly drive customer lifetime value.
In this realm of predictive analytics, structured or unstructured data alone do not provide enough customer input data to arrive at meaningful and accurate predictions. Both have to be combined together with an application of statistics to arrive at critical customer behavior predictions that increase customer lifetime value.
emStream's deep expertise in structured data analytics through use of tools like R, WEKA and analysis of unstructured data through Natural Language Processing put us in a sweet spot to map customer profile, transaction and sentiment data across all channels and use these on large scale HADOOP clusters to predict customer behavior real time.