Presentation by Meshlabs at Zensar #TechShowcase - An iSPIRT ProductNation initiative.

MeshLabs Text Analytics

©  2013  MeshLabs  So0ware  Private  Limited  

Confiden<al  

About US

2  

Featured Customers:

Provider  of  text  analy<cs  so0ware  products  

Informa<on  Management  |  Customer  Experience  Management    |  Business  Intelligence  |  Regulatory  Compliance  

ü  On-­‐premise  ü  SaaS  ü  API  

ü  Unified  Content  Access  ü  En<ty  Extrac<on  /  Tagging  ü  Categoriza<on  ü  Summariza<on  ü  Recommenda<on  ü  Faceted  Search  ü  Sen<ment  Analysis  ü  Dashboard  &  Repor<ng  

Text, Text, Everywhere…

3  

4  

Too much volume and variety   Missed  Opportuni1es  

Product Managers Customer Insight Managers

Research Analysts Customer Care Reps

Sales & Marketing Leaders HR Leaders

Senior Executives  

Cost / Quality concerns over manual methods  Current BI tools won’t work  

Structured data only and too complicated  

And Not a Single Insight.

Multiple Channels, Sources and Types

Limited Analysis, Ad hoc, Scalability

Issues  

Topline and Bottom-line Impact  

Text Analytics

5  

Linguis<cs  

Sta<s<cs  

Seman<cs  powerful  technology  to  automa<cally…  

Ingest  all  text  data/content  

Extract  valuable  assets  

Deliver  ac<onable  insights  

1  

2  

3  

How it Works

6  

1   v  Connectors to Enterprise Content Stores, Facebook, Twitter etc.

v  Crawlers for getting data from websites v  Upload files & documents – Excel, Word, PDF etc.

2   Process your data – Extract entities, classify, cluster, and score sentiment

v  NLP – Natural Language Processing v  Taxonomies & Custom Ontologies v  Machine Learning

3   Analyze output - dashboards, reports, workflows, and alerts

v  Dashboards v  Charts & Reports v  Exports

Gather your data – Text (Unstructured) and Structured

Key Use Cases

7  

Informa<on  Extrac<on  

“How  do  I  extract  key  informa<on  from  CRM  Notes  to  

predict  cross-­‐sell  &  up-­‐sell  opportuni<es”  

Opinion  Mining  

“  How  do  I  gain  ac<onable  

insights  from  market  &  customer    interac<ons  

across  channels?  ”  

Auto-­‐Categoriza<on  

“  As  a  retailer,  how  do  I  display  

categorized    lis<ngs  in  the  most  efficient  manner?  “  

Intelligent  Agents  

“  With  so  much  

informa<on  overload,  how  do  I  transform  

the  effec<veness  

of  my  knowledge  workers?  “  

Customer Testimonial

8  

“We  partnered  with  MeshLabs  because  of  their  unique  ability  to  connect  to  and  integrate  all  types  of  data  and  content  from  our  communi<es.  This  allows  us  to  bring  game  changing  analy<cs  and  repor<ng  to  our  clients  enabling  them  to  discover  new  insights  to  refine  messaging,  cra0  an  innova<on  strategy,  and  improve  customer  loyalty.”  

THOMAS  FINKLE  CEO,  Think  Passenger,  Inc.  

Passenger  is  a  leader  in  providing  Market  Research  Online  Communi1es  PlaKorm  

Our Product – eZi CORE™ Text Analytics Engine

9  

MeshLabs eZi CORE ™

eZi Semantic Search ™

eZi Reco ™

eZi Connectors ™ and CrawlersMicrosoft SharePoint, Outlook, Alfresco

Enterprise ContentWeb Content

eZi Sentiment Analyzer ™

Entity Extractor

POS Tagging Classifier Clustering Rules

EngineInference /Reasoner

Unified Semantic Index / Triple Store

Search Interface Dashboards APIs

Custom Solutions

•  On-­‐Premise  •  SaaS  •  API  

Core Capabilities

10  

ü   Data  Acquisi<on  and  Inges<on  ü   Text  Prepara<on      ü   Named  En<ty  Extrac<on  ü   Auto-­‐Categoriza<on  ü   Feature  Extrac<on  ü   Sen<ment  Analysis  ü   Summariza<on  ü   Recommenda<on  ü   Faceted  Search  ü   Dashboard  &  Repor<ng  

Data Acquisition and Ingestion

11  

•  File  System  •  SharePoint  •  Alfresco  •  Web  Crawler  

•  TwiUer  •  Facebook  •  Blogs  •  YouTube  •  Discussion  Forums  

•  Yahoo  Answers  •  Hadoop  File  System  (S3,  

HDFS  etc.)  •  Databases  (any  JDBC  

compliant  database)    

12  

Out-of-the-box Taxonomies

Airline  Industry   Automobile  Industry   Banking  

Company  -­‐  Industry  

Classifica<on  

Computers  and  Laptops  

Corporate  Social  Responsibility   Cosme<cs   Customer  Service  

-­‐  Generic  

Hotels  Human  Resources  

-­‐  Voice  of  Employee  

Product-­‐Category  Classifica<on   Real  Estate  

Retail   Smart  Phones  and  Tablets   Telecom   Travel  

13  

Dashboards & Reports

Sentiment Analysis

14  

•  Feature-­‐Based  Sen<ment  Analysis  supported  •  Lexicon  based  analysis  

§  Per-­‐domain  lexicon  supported  •  Uses  deep  parsing  to    

§  Iden<fy  features  §  Associa<on  of  nega<on  and  suppor<ng  words  

•  Mul<ple  levels  of  sen<ment  scoring  supported  •  Sen<ment  Analysis  done  at  sentence  fragment  scope  •  Weighted  rollup  of  sen<ment  score  provides  overall  

view  

Feature Detection

15  

•  Features:  Extrac<on  of  Context  Relevant  Nouns  /  Noun  Phrases    ü  Noun  Phrase  Extrac<on  ü  Deep  Parsing  and  Lexical  Chaining  ü  Sen<ment  Scoring    at  Feature-­‐level  

“The  coffee  was  bad,  but  the  sandwich  was  good.”    •  Featureless  Sen<ment  Score  –  Neutral  •  Featured-­‐based:  

ü  Overall  Sen<ment  –  Neutral  ü  Coffee  –  Nega<ve  ü  Sandwich  -­‐  Posi<ve  

Contact Us

16  

[email protected]  www.meshlabsinc.com  

@meshlabs  linkedin.com/company/meshlabs  facebook.com/meshlabs  

USA:  1-­‐602-­‐617-­‐9370    |    India:  91-­‐9986004572