Top Banner

Click here to load reader

Veda Semantics - introduction document

Aug 11, 2014



Big Data analytics, social media analytics, text analytics, unstructured data analytics... call it what you may, we see ourselves as experts in text mining and have products and services that provide insights from various kinds of unstructured data. Already recognized by Gartner for our expertise, we are passionate about what we do and have also filed patents for some innovative approaches we have used.

  • Veda Semantics Building intelligence through semantics Text Analytic s Text Analytics Ontology Building Context Analysis Sentiment Analysis Machine Learning
  • Introduction Natural Language Processing use cases and Discovery product Text analytics use cases, Prism and Txt products Examples of Veda projects and capabilities 2
  • About Semantic technology 3 Semantic technology is a language processing framework that helps make sense of unstructured data lying in documents such as PDFs, Word documents, Emails Highlights of Semantic Technology Process data in a manner similar to how the human mind understands data Example: Joe works for XYZ Corp A semantic framework through its linguistic processing models understands that Joe is a name, works is a verb and XYZ Corp is an organization. Extract concepts and sentiments from any sentence Example: Joe loves the seats of the Honda Civic Semantic frameworks combined with lexicon engines auto classify the above sentence as a positive sentiment for the seat of a car. Establish linkages between data across heterogeneous sources Example: Joe works for XYZ Corp (Document 1) ; Joe loves the seats of the Honda Civic (Document 2). Joe XYZSeats PDF Word Emails Social Media Semantic Engine Link Analysis Structured Data for search Sentiments
  • A semantic technology company Veda Semantics has expertise in both Natural language processing and Statistical text mining techniques for Big Data scenarios About Veda Semantics Experienced team Key members of technology team each have over a decades worth of experience each in Semantic technology 4 Reputed leadership Mr. V. Srinivasan Chairman of the group with over 30 years of experience in Banking and IT, and ex-global MD & CEO of 3i Infotech Ltd. Mr. Rajat Kumar - CEO who is a Wharton and McKinsey alum, with experience across diverse geographies and functions. Focus Areas Text Analytics through use of Statistical Algorithms with a NLP overlay Sentiment Analysis through NLP and Lexicon Engines
  • Veda Semantics has been recognized by Gartner in two separate reports (Whos Who of Text Analytics, September 2012, and Report on BI platforms in Asia, January 2014) About Veda Semantics 5
  • Veda Difference 6 Entity Extraction Recognizes people, events, date, organizations automatically Veda Difference Unlike traditional search which is based on keywords, Vedas technology backbone is a combination of advanced statistical and language processing algorithms then help not only understand data contextually but allow a user to FIND what they are looking for with a very high level of relevance and KNOW what they need to look for if they have no clue where to start Vedas sentiment engine deep dives to identify sentiments at clause level that translates into actionable insights at a product attribute level Key Features Document Classification Groups similar documents together for easier search Concept Extraction &Linkages Automatically extracts key concepts from text, classifies them and associates related concepts across documents Hadoop Ability to process large volume data over commodity hardware in parallel Sentiment Analysis at Attribute Level Extracts sentiments and attaches them to attribute of a product Data Correlation Gets correlation between related terms
  • Veda Difference 7 Sentiment Analysis companies Veda Prism Text Analytics Companies Keyword Search Bag of Words Document Classification Visual Entity Segregation Related term and action association Sentence Based Sentiments Connecting to multiple sources Response Dashboards Clause Level Sentiment Easy taxonomy creation Competition monitoring at attribute level Veda Discovery
  • Veda Technology Stack Proprietary Linguistic Processing Capabilities Vedas linguistic processing capabilities including Entity Extraction, Anaphora Resolution, Clause Level Identification are proprietary. The core technology has evolved with years of R&D thereby giving it a high level of accuracy Ability to process unstructured data in multiple formats Connectors to various sources including PDFs, Word docs, Excel, Outlook allow processing of data from heterogeneous sources and convert it into structured data stores Patent for visual entity segregation Veda has filed a patent for visual entity extraction. Even without supporting context, the engine can pull out relevant information based on the document structure. Veda is also filing a patent for the proprietary clause based sentiment technology User Interface allows sophisticated charting and drill down to get to the bottom of things Use of intuitive charts and other advanced front end charting technologies allow for data visualization at a whole new level Robust Architecture and Seamless Integration Vedas robust architecture of which Hadoop and Storm are a key component allow for real time and batch processing of millions of records. Our technology stack can be readily integrated with any application
  • Introduction Natural Language Processing use cases and Discovery product Text analytics use cases, Prism and Txt products Examples of Veda projects and capabilities 9
  • Sample use cases NLP & Sentiments 10 Reputation Management Product Improvement Areas Monitoring large volume social media feeds and converting them into actionable insights Provides connectors to social media to monitor feeds across Twitter, Facebook, blogs, internal emails and throws sentiments for each user, location, etc. Tracking sentiment for products, competitors. Monitoring online reputation and responding to negative publicity Vedas platform allows real time monitoring of social feeds to understand swings in customer sentiment and allows companies to act on them immediately Monitor customer feedback, internal feedback for products across email, chats, forums to understand customer feedback Details product attribute level sentiment and action terms that makes feedback highly actionable Medical Development Monitoring online posts of patients to check for possible adverse psychological reaction to test drugs Advanced sentiment engine can track trends over time, allowing a comparison to be made between pre and post drug use NLP can be used to track employee suggestions, motivation levels and use as an input in product launch or project success predictions Employee suggestions can be tracked deeply and in aggregate in a mater of a few clicks. Easy hierarchy building allows a top level view of what critical areas require immediate management attention Sentiment Tracking Employee Suggestions Industry Challenge Veda Difference
  • Benefits of using Veda in Voice of Customer 11 Clause level sentiment identification gives sentiments at a attribute level that gets grouped at a category level The Veda Semantics platforms have an ability to ingest various forms of data, including txt and word documents, PDFs and Excel Response dashboard allows users to instantly respond to negative comments or sentiments Customer Support for a large FMCG Currently, it is difficult to get a high level summary into the areas of poor service. The Veda Semantics engine gets feedback data from multiple sources such as Email, Chat, etc. which is unstructured and structures it in intuitive categories. Veda Edge Sentiment Time Series allows users to look at and deep dive into sentiments for specific time periods, across locations A list of top influencers allows a check into critical people who need to be addressed on social media Veda Engine Competition analysis allows for side-by-side comparison for competitor products
  • Features of Veda Discovery: Our flagship product for sentiments Real Time Monitoring Sentiment Scoring/Averaging Competition Mapping Time Series/ Influencer Analysis Social Responses Depth (Clause Level) Veda provides real time monitoring of social media feeds for real time insights and responses Allows trends over time to be considered with the ability to deep dive into a particular time period. Displays top influencers and sentiments around them A critical benefit of using the Discovery product that traditional sentiment engines tools do not have is Clause level identification and mapping. Veda can look deep into a sentence to determine what a sentence is talking about Sentiments are scored from a scale of -5 to +5. Do not like is less negative than hate, and amazing is more positive than great. Vedas algorithms work to provide average scores across reviews for each aspect being considered. Comments indicating Intent to buy are highlighted separately. Side by side monitoring display of competition. Can be done not only at the overall brand level, but also at the attribute level (e.g. perception about own vs. competitor price / quality / looks, etc.) Respond to social messages through the dashboard itself
  • Veda Discovery Sentiment analysis 13 Domain Example sentence Sentiment Other Engines Sentiment - Veda Electronics I love the screen of the phone. Positive Sentence Positive sentence, assigns it a score, and link the positiveness to screen Hospitality I love the room, but hate the service! Neutral Sentence Positive for room Negative for service Assigns marks to each attribut