Closed loop with Computer Linguistics

0 %
100 %
Information about Closed loop with Computer Linguistics

Published on September 28, 2015

Author: jurajschick


1. scopeKM Knowledge Management „Closed Loop“ with Computer Linguistics Discover, analyze, and combine customer experiences and successfully reflect them into strategy Presentation Juraj Schick scopeKM GmbH, Zürich 05.12.2014 1

2. scopeKM Knowledge Management Agenda 05.12.2014 2 1. Challenges and Opportunities 2. The view of the things behind: Identify, apply and deepen the knowledge

3. scopeKM Knowledge Management Purpose of the text analysis €Recognize relevance, assess the situation and act purposefully € The solution uses the Natural Language Processing (NLP) to „understand” facts, sentiments and their context. € The barriers of otherwise isolated data silos can be overcome and joint analysis of structured and unstructured information (text flow) enabled. € The relevant information will be detected, extracted, and semantically enriched with company-specific metadata. Purpose: creation of customer or product- specific documentation to ensure the end-to-end management using workflow engine. € NLP allows to extract information about interesting entities and their relationships in order to derive precise and relevant RDF-triples (Subject <> Predicate <> Object). In this way, the causes of events can be determined – the answer to the question of "why something happens" – in order then to act targeted and appropriate to the situation. 05.12.2014 3

4. scopeKM Knowledge Management The new challenges €The challenges of technological change € Steadily increasing amount of data (terabyte/d. +). € Complexity of relations with disproportionately rising costs (many customer channels) and inhibited learning strategies (lack of closed loop). € Changing communication requirements (from online to mobile). € Data silos (file server, CMS, CRM/call center, mail server, etc.) with unstructured information (80% + of all data in the enterprise), which are not usable with conventional methods. € Non - or little structured knowledge processes as a cross-cutting functions. 05.12.2014 4

5. scopeKM Knowledge Management Understand the meaning and context, act purposefully €The large and growing data volume and complexity can be overcome only by (partial) automation of services € Opening up of all internal and external sources, linking the structured (BI) with the non-structured data (text flow). € Closed Loop: automatically understand and analyze, communicate and kick off stimulate, act as well as track and learn. € Only the answers assessed by a confidence factor as successful and which are available in the context of the task are presented . € Machine learning systems have the ability to •understand‚ and to •interpret‚ the issue in a broader context and ƒ if necessary ƒ to ask questions in order to clarify the task. 05.12.2014 5

6. scopeKM Knowledge Management „Closed Loop“ - CEM Decisions Analysis Tracking/ Learning Actions CEM* comprehensive & prospective collaborative or automatically in real-time & informed pervasively & continuously * CEM Customer Experience Management 05.12.2014 6 Discover, analyze, combine and reuse

7. scopeKM Knowledge Management The new chances € The early discovering of new trends and the correct action towards the changed customer behavior. € Industrialization of processes and benchmarking (Best Practice incl. KPI). € Linking macro with micro-analyses for improved understanding of natural and human-caused risks. € Macro-analyses: Identification of trends and patterns for the determination of the strategy and methodology (discover new insights about the opportunities, behavior, performance, etc.) € Micro-analyses: Knowledge about single customers or support of single activities, like diagnoses, assessments and consultations. € Ability to discover operational challenges and opportunities in real time and to answer them proactively. 05.12.2014 7

8. scopeKM Knowledge Management Agenda 05.12.2014 8 1. Challenges and Opportunities 2. The view of the things behind: Identify, apply and deepen the knowledge

9. scopeKM Knowledge Management With NLP, different linguistic representation levels are processed sequentially: € Tokenization: is the process of breaking a stream of text up into words, phrases, symbols, or other meaningful elements called tokens. € Morphology: is the identification, analysis, and description of the structure of a given language's morphemes and other linguistic units, such as root words, affixes, parts of speech, etc. € Syntactic analysis/parsing: is the process of analyzing a string of symbols, conforming to the rules of formal grammar, recognizing the basic scheme of the source sentence (subject, predicate, secondary elements), and processing Part-of-speech-Tagging. € Semantic analysis: meaning is assigned in different single steps to the sentences or its parts. NLP Natural Language Processing €To understand and to classify the facts and the sentiments, the language – the written text - must be understood. This is accomplished with the NLP. 05.12.2014 9

10. scopeKM Knowledge Management Semantics as a Solution € Semantics are based on relationships between data and are therefore ideal tool for linking and searching for complex structured and unstructured data with the standard query SPARQL. € Example insurance: Inquiry "All beneficiaries of health insurance, earning more than $ 100,000 in 2010 and living in Atlanta, GA“ indicates the combination of data on income, geography and time. € With semantic solutions, all sources are linked and isolated data silos avoided.

11. scopeKM Knowledge Management Luxid ® : Components of the Solution 05.12.2014 11 Luxid ® Annotation Factory as a pipeline for the NLP € Extraction of Topics, Entities and Relationships from the text € Categorization of documents and their clustering € Extraction engine for syntax, statistics, taxonomy, etc. € definition of rules for clustering, displaying, etc. Luxid ® Skill Cartridge Library for different applications or application areas, such as taxonomy/ontology, text mining, opinion mining or marketing, finance, medicine, business administration, pharmacy and chemistry, etc. Luxid ® Content Enrichment Studio for customizations of existing or development of completely new applications.

12. scopeKM Knowledge Management The triple – the logical statement about resources € A triple is a way of encoding information about objects in the RDF (resource description framework). The triple is a basic form of the ontology. € The triple represents an assertion of subject and object in relationship with each other. Relations are directed from the subject to the object and concretized with the predicate. Triples, which relate to the same subjects or objects, form a semantic network. Subject Predicate Object ZKB grants mortgage Mortgages require equity The mortgage has an interest rate ZKB is a bank In the example table each line forms a triple: 05.12.2014 12 E.g.: analysis of the inferences of aspects without formal representation: Interest rate= f(equity)

13. scopeKM Knowledge Management € The triples as statements about objects (resources or entities) are stored in knowledge bases. Then the knowledge bases can be integrated seamlessly into the workflow applications and analysis tools of the end users. € Luxid „ extracts business information from unstructured texts, structured it in the triples, which can subsequently be queried, visualized and analyzed. For example: € Which business relations are there between a potential partner and the competitors? € How are our products judged in the media, compared to analogous products of our competitors? € Which customers comment on what concerns about the product Abc in relation to a competitor's product? Linking all data – company-wide 05.12.2014 13

14. scopeKM Knowledge Management Notations and their hierarchy 05.12.2014 14 Descriptors Annotations Metadata Categories Concepts Terms Relationships Entities € Annotations: Descriptors created at text level (concepts) or document level (terms) € Concepts: Normalized forms created in text analyses to which terms and expressions are mapped € Entities: Concepts that represent self-explanatory units of information, possibly with attached attributes (e.g. the expression United States… is mapped to entity •USA…) € Relationship: Concepts that link several entities at the sentence level, possibly with assigned role for corresponding entity and attached attributes (•Company A acquired Company B in January 2014 …; Company A: subject, Company B: object, in January 2014: the context of the relationship € Terms: The most important words or groups of words in a document, which are used for clustering or categorization € Categories: Classification of documents according to the concepts they contain (taxonomy/ontology).

15. scopeKM Knowledge Management Which domain metadata ? € People names: Customers, Collaborators, Links to the directory € Organisations: Subsidiaries, Departments, Suppliers, Competitors, Partners € Internal References: Project, Contract, Customer, Geography, Market Segment, etc. € Product- und Service names: Product taxonomy, Accessories, Options, etc. € Enterprise Terminology and Categories: Technical vocabulary, Document categories 05.12.2014 15

16. scopeKM Knowledge Management Automatic extraction of metadata On May 10th Microsoft bought Skype for $8 Billion. Terms Num ProperPrep Verb Proper Prep U NProp Card CompanyPrep Action Company Prep Monetary Expr.Date Entities Relations Acquisition Purchaser Microsoft Target Skype Amount $8 billion Date May 10th On May 10th Microsoft bought Skype for $8 Billion. Roles Attributes 05.12.2014 16

17. scopeKM Knowledge Management T T T T T T T T T T T T T T T T T T T T T T T T T T T T T TT TTTT TTT TTTT Creation of the taxonomy/ontology Tagging of documents Definition of the triples Taxonomy/ontology to the document-layer Application of the taxonomy/ontology Seamless synchronization: transparency for the end- user Luxid® Webstudio: automatic ontology management + + € Preview shows the annotations in real-time € NLP driven proposals for improving the thesaurus 05.12.2014 17

18. scopeKM Knowledge Management Opinion Mining € With the Opinion Mining Skill Cartridge content created by users will be analyzed in order to identify synthetic views in them – with all Pros and Contras – and to evaluate the tell-tale signs within the information flows. € The subjective information-bearing linguistic character in the text is recognized and used for identification and qualification of evaluative expressions. € The result of data mining are then valuable and non-obvious Insights. € The sentiment analysis represents a part of the overall Opinion Mining solution. 05.12.2014 18

19. scopeKM Knowledge Management Key Elements of Opinion Mining €Fine granularity of evaluation result is achieved by determining polarity and intensity of the statement. 05.12.2014 19 OpinionSource Target Negative Positive Weak Strong… Affect Polarity Intensity Judgement Evaluation type …

20. scopeKM Knowledge Management The Screenshot shows, that positive and negative opinions are expressed on a TV product and ist features from BrandX, with more or less intensity. The Proof zone provides the full sentence for context. Sentiment analysis 05.12.2014 20

Add a comment

Related pages

Control theory - Wikipedia, the free encyclopedia

In a closed-loop control system, ... In complicated systems this can require computer-assisted calculation capabilities, and cannot always ensure robustness.
Read more

Closed loop - Wikipedia, the free encyclopedia

Closed loop may refer to: ... a commonly used closed-loop controller; Closed ecological system not relying on matter exchange outside of the system, ...
Read more

Closed-loop System and Closed-loop Control Systems

Closed-loop Systems. In the previous tutorial we saw that systems in which the output quantity has no effect upon the input to the control process are ...
Read more

How to Tune Closed Loop and Open Loop o2 -

Closed loop is when the ECU does refer to the o2 sensor for ... Once you have fully tuned the fuel map to be as close as you can to your target A/F ...
Read more

Closed-Loop Recycled Content | Dell

How Dell does closed-loop recycling ... Meanwhile, other materials from the computers are similarly recycled and separated, ...
Read more

Closed-Loop Supply Chain Network Design with Recovery of ...

... Computational Intelligence Computational Linguistics Computer Simulation Computer Vision & Image ... Closed-loop supply chain management is ...
Read more

BBC - GCSE Bitesize: Open- and closed-loop systems

Open- and closed-loop systems Open-loop systems. Open-loop systems are set up to achieve desired results, but there's no way of checking if the results ...
Read more

Modeling Closed Loop Supply Chain Systems: Business and ...

Modeling Closed Loop ... Intelligence Computational Linguistics Computer Simulation Computer Vision ... Closed loops consist of two integrated ...
Read more

Validation and test of a closed-loop tele-rehabilitation ...

Validation and test of a closed-loop tele-rehabilitation system based on functional electrical stimulation and computer ... closed-loop tele ...
Read more

Closed Loop Control Logic - How is Closed Loop Control ...

It is Closed Loop Control Logic. ... (linguistics) CL: ... Closed Loop Control Logic; Closed Loop Dynamic Stability Test;
Read more