Silvia Calegari, Marco Comerio, Andrea Maurino,

Similar documents
KDIR th International Conference on Knowledge Discovery and Information Retrieval

A logistic regression model for Semantic Web service matchmaking

An Approach to Event-based Dynamic User Profiling in Social Media

Effective Partner Compliance Strategies

An Introduction to Use-Case Modeling

SPECIAL INSTRUCTIONS TO BIDDERS FOR PREPARATION OF THE PROPOSAL

Improving Web Service Clustering through Ontology Learning and Context Awareness

On utility of temporal embeddings for skill matching. Manisha Verma, PhD student, UCL Nathan Francis, NJFSearch

Policy for integrity and marketing activities. Latest update: 21 May General

Strategic Management of Oil and Gas Assets and Companies

OntoNaviERP: Ontology-supported Navigation in ERP Software Documentation

Introduction to Recommendation Engines

Predictive analytics [Page 105]

Contract notice & conditions for participation AQuA Affiliate/s: Motivational Interviewing

Microsoft Dynamics GP. Purchase Order Processing

Development and Evaluation of An Employee Performance Appraisal Insight Report Generator

Artificial Intelligence Exploration (AIE) Opportunity DARPA-PA Automating Scientific Knowledge Extraction (ASKE)

CHAPTER 2 PROBLEM STATEMENT

Hilltoppers Swimming Online Registration Guide

Automatic Tagging and Categorisation: Improving knowledge management and retrieval

PubRec: Recommending Publications Based On Publicly Available Meta-Data

IR Research: Systems, Interaction, Evaluation and Theories

XPLODIV: An Exploitation-Exploration Aware Diversification Approach for Recommender Systems

License-aware Service Selection

General Terms and Conditions

A Business Rules Perspective on a Standard Rules Language. Gary Hallmark April 28, 2005

UMBC CMSC 461 Database Management Systems. Course Project

REASONING ABOUT CUSTOMER NEEDS IN MULTI-SUPPLIER ICT SERVICE BUNDLES USING DECISION MODELS

Quality Assessment Method for Software Development Process Document based on Software Document Characteristics Metric

CHAPTER VI FUZZY MODEL

APPLICATION FOR PROFESSIONAL REGISTRATION AT CHARTERED ENGINEER LEVEL BY THE ROYAL INSTITUTE OF NAVIGATION ROUTE

Semantic Information Management for Pervasive Computing

Comments on Key Performance Indicators ( KPI ) Matrix and Statistical Validity

[Type text] BACKGROUND:

EndCustomer Monitor 4.0

A Fuzzy Multiple Attribute Decision Making Model for Benefit-Cost Analysis with Qualitative and Quantitative Attributes

REQUIREMENTS SPECIFICATION MAPPING (RSM) CCBDA Cross Industry Catalogue Process v2.0

CHECKLIST Quality Assessment for Final Evaluation Reports. Official(s) managing the evaluation:

Ontology-Based Model of Law Retrieval System for R&D Projects

CHAPTER 5 SUMMARY AND CONCLUSIONS

ASSET MANAGEMENT SYSTEMS ANALYSIS

An Automated Decision Support System to Assist with Project Planning, Program Management and Work Flow Analysis of an Enterprise

Keywords:Traceability, Requirement, Requirement Traceability techniques, Traceability Links, Information Retrieval.

Models in Engineering Glossary

PARTNERSHIP EXPRESSION OF INTEREST FORM

Big Data in Emergency Informatics Social media data perspective. Rajendra Akerkar

Bloomberg Ariba Network. Online Invoice Guide

Contoh Pengembangan SIM

GrammAds: Keyword and Ad Creative Generator for Online Advertising Campaigns

Using Data Analytics to Detect Fraud

Seminar in E-Business & Recommender Systems University of Fribourg, Department of Informatics

QUALITY ASSESSMENT FORM 1

InCites Benchmarking & Analytics

BIM Contracts & Risk

Predicting user rating for Yelp businesses leveraging user similarity

IM S5028. Architecture for Analytical CRM. Architecture for Analytical CRM. Customer Analytics. Data Mining for CRM: an overview.

Product Release Notes for ACTIVEWorks Endurance May 4, 2017

Terms of Trade GENERAL SELLER S OBLIGATIONS. Terms of Trade web-parts.com, 7. March is owned by

Job Board - A Web Based Scheduler

Context-aware recommendation

PROFIT INFORMATION. ABA Service Membership A Closer Connection to Your Customers EXPOSURE ACCESS GROWTH PROFIT SAVINGS VALUE ROI ROI ACCESS GROWTH

OPplus. Vertical Solution. Manual TRIAL BALANCE & AFFILIATION. Microsoft Dynamics NAV all versions. for. Certified For Microsoft Dynamics NAV

Prescription of qualifications. ARB Criteria at Parts 1, 2 and 3

Evaluating Workflow Trust using Hidden Markov Modeling and Provenance Data

Evaluating Enterprise Architectures through Executable Models

Improving the Response Time of an Isolated Service by using GSSN

Centre for Computational Finance and Economic Agents WP Working Paper Series. Biliana Alexandrova-Kabadjova, Edward Tsang and Andreas Krause

Security deposits and container deposits

APS Bank plc Data Privacy Policy

MICROS inmotion Hosting Installation Guide

HKT Financial Services (IA) Limited Privacy Statement

OntosCAI Competitive Affairs/Intelligence Analyze, Monitor, Understand your Competitive Environment

ICIMTR International Conference on Innovation, Management and Technology Research, Malaysia, September, 2013

Recasting the context in relevance feedback

FCTA th International Conference on Fuzzy Computation Theory and Applications

Operational Application of Targeted Data Analysis. Eric Chasin, NTELX Agenda. What we do?

Invoice Status Service

ARIS PROCESS PERFORMANCE MANAGER

How to Build, Implement, and Use an Architecture Metamodel

Ask the Expert SAS Text Miner: Getting Started. Presenter: Twanda Baker Senior Associate Systems Engineer SAS Customer Loyalty Team

frequently asked questions

An Analysis Framework for Content-based Job Recommendation. Author(s) Guo, Xingsheng; Jerbi, Houssem; O'Mahony, Michael P.

PayWell Moldova 2017 Salary and Benefits Survey

Dynamic Generation of Personalized Product Bundles in Enterprise Networks

Parameter Estimation for Diagnosis and Optimization in Power Plants

Pricing Website Packages Type Description Price

Risk assessment checklist - Purchasing cycle

Karmann Mills 1, Anthony Hickey 1, Alexander Tropsha 2

This lecture will cover. Quantitative Research When to use Quantitative Research Quantitative Research Methods

GOVERNMENT OF KENYA EXPRESSION OF INTREST NO. ICTA/EOI/01/ FOR

Enterprise Applications powered by

A New Approach Towards Intelligent Analysis for Competitive Intelligence *

Onli Onl ne R egi ne R egi trati a on, Accom on, Accom od m a od ti a on, Ma

Manufacturing Capability Inference and Supplier Classification Based on a Formal Thesaurus

Genie De Banca: Go Beyond Digital & Multiply Revenues. Ride the digital wave with innovative digital banking solution and open up new revenue streams

DEADLINE: NOVEMBER 20TH

The Customer Is Always Right: Analyzing Existing Market Feedback to Improve TVs

Compliance Management basierend auf Gesetzesformalisierungen Das REALM-Projekt

Anatomy of the CIPS Database of Procurement Professionals

Advisory Services in Payments

Transcription:

A Semantic and Information Retrieval based Approach to Service Contract Selection Silvia Calegari, Marco Comerio, Andrea Maurino, Emanuele Panzeri, and Gabriella Pasi Department of Informatics, Systems and Communication (DISCo) University of Milano-Bicocca {calegari,comerio,maurino,panzeri,pasi}@disco.unimib.it 1

Contents Problem Definition Motivation, Background and Contributions The Semantic and IR based Approach Multi-constraint query formulation Filtering and query evaluation Experimental Results Conclusions and Future Works 2 2

Service Contract A Service Contract represents the agreement between a service provider and potential service consumers to use a specific service under given conditions. Beyond the description of service functionalities, a service contract is composed by contractual terms on: Quality of Service (e.g., response time and availability); Legal Terms (e.g., limitation of liability and copyrights), Intellectual Rights (e.g., denying composition), Business Terms (e.g., payment and tax). 3 3

Service Contract Selection For each service, multiple service contracts are available. Each service contract can be offered to specific user categories; Each user category is associated with specific affiliation conditions. Service Contract Selection: identify the service contracts that better fulfill the constraints on contractual terms explicitly specified by the user, and/or implicitly inferred from user information. 4 4

Service Contract Selection How to automatically ti select the best among a set of functional-equivalent services? 5

Contents Problem Definition Motivation, Background and Contributions The Semantic and IR based Approach Multi-constraint query formulation Filtering and query evaluation Experimental Results Conclusions and Future Works 6 6

Motivation & Background An approach to automatic service contract selection should cover the following characteristics: expressivity as the possibility to evaluate qualitative contractual terms by means of logical expressions on ontology values, and quantitative contractual terms by mean of expressions including ranges and inequalities; extensibility as the possibility to customize evaluation functions; flexibility as the possibility to perform evaluation in case of incomplete specifications. 7 7

Motivation & Background In ICSOC 2009, we proposed an hybrid approach to service contract selection that combines logic-based and algorithmic techniques; offers high levels of expressivity, extensibility and flexibility. The approach has been implemented by the Policy Matchmaker and Ranker (PoliMaR) framework that operates on service contracts defined according to the Policy Centered Metamodel (PCM). Beyond performance problems, the approach and the framework presented the following limitations: i i no support for the formulation of user requests; no support for the evaluation of user category affiliations. PoliMaR is available at: http://sourceforge.net/projects/polimar/ 8

Contribution A new approach to service contract selection based on: the exploitation of preferences explicitly specified by the user, and implicitly inferred from user information; the use of both semantic-based and information retrieval (IR) techniques to filter and rank service contracts. 9 9

Contents Problem Definition Motivation, Background and Contributions The Semantic and IR based Approach Multi-constraint query formulation Filtering and query evaluation Experimental Results Conclusions and Future Works 10 10

The Proposed Approach Set-up time Run time 11 11

Registration At set-up time, the user: selects a pre-defined profile providing information on generic user characteristics; inserts personal information; specifies preferences. 12 12

Registration - Example Pre-defined User Profile I am an English speaker. I am able to use my mobile phone and to frequently access my email account. I am a VAT owner. User Profile PCM Name: Mary Brown Address: London, Oxford street Age: 45 years old Job: IT Researcher Language: English Info. Channel: email, phone call VAT owner: yes Preferences: [secure, cheap] 13

Query Formulation The user selects a pre-defined query presented as a textual description of both precise (e.g., insurance = blanket) and flexible constraints (e.g. price = at most 40 ). The user personalizes the query by modifying the pre-defined constraints, and/or by adding further constraints as short textual descriptions. A query expansion process is applied to add further constraints from user profile and user history. 14

Query Formulation An Example Pre-defined User Query I need to perform the transportation of a valuable good. I am looking for a fast delivery service having a blanket insurance. Personalized Query Delivery in at most 24 hours. Price at most equal to 40. I would like to receive traceability information on the transportation User Profile Name: Mary Brown Address: London, Oxford street Age: 45 years old Job: IT Researcher Language: English Info. Channel: email, phone call VAT owner: yes Preferences: [secure, cheap] PCM 15

Filtering Each service contract is offered to one or more user categories that are defined by a set of affiliation conditions (e.g., user age, VAT owner). A user is associated with a category if and only if all the conditions are respected. Service contracts are filtered complying to the user category affiliations that are determined analyzing user profile and user history. The result is a set of filtered service contracts. 16 16

Filtering An Example User History Provider Contract Category Conditions Provider A pay-flex Business Plus VAT owner, email account, 30 shipments Provider A high-trace Business One VAT owner, email account Provider A secure Business One VAT owner, email account Provider B fast-plus Silver User 20 shipments Provider B fast Bronze User 10 shipments Provider B cheap Senior User 65 years old 17

Query Evaluation The multi-constraint query is evaluated against the filtered service contracts. A ranked list of service contracts is returned to the user. Query constraints and contractual terms are expressed by both specific data and textual descriptions. Different evaluation functions are needed to evaluate the matching degrees between constraints and contractual terms. An aggregation function is used to compute the overall service contract score. 18 18

Query Evaluation Evaluation Functions Constraints on numeric data values: constraints expressed as fuzzy subsets of the attribute domains; evaluation perfomed by means of parametric linear membership functions; e.g., A membership function for price at most 40 constraint. Concept-based constraints: constraints expressed on concepts defined in ontologies; evaluation performed on the basis of the semantic distances between required and offered values. 19 19

Query Evaluation Evaluation Functions Keyword-based constraints: information retrieval techniques are used to extract t keyword from contractual terms expressed in plain texts; evaluation performed using the Vector Space Model that represents each set of keywords as vectors and supports the evaluation of the similarity between two vectors using a vector distance (e.g., the Cosine similarity). 20 20

Query Evaluation Evaluation Functions The overall service contract score (namely, Degree of Match DoM) ) is computed using the following aggregation function: nc r r [ (, )] + (, ) 1 (, ) = CF sc q CosSim sc q i= i DoM sc q nc + 1 nc = number of query constraints; CF i = evaluation of constraint i; CosSim(sc,q) = evaluation performed using Cosine Similarity on sc (i.e., keyword vector associated with the contract) and q (i.e., keyword vector associated with the query). 21 21

Query Evaluation An Example Multi-constraint Query PCM Language: g English Info. Channel: email, phone call Hours to Delivery: at most 24 hours Price: at most 40 Insurance: Blanket Pay. Methods: credit card, electronic transfer Preferences: [secure, cheap, traceability] CONTRACT FAST-PLUS PCM Info. Channel: SMS Hours to Delivery: 12-24 hours Price: 40 Insurance: Fire and Theft Pay. Methods: credit card 0.0 1.0 1.0 0.33 0.5 Description: [fast, 22 traceability, english] 0.50 22

Contents Problem Definition Motivation, Background and Contributions The Semantic and IR based Approach Multi-constraint query formulation Filtering and query evaluation Experimental Results Conclusions and Future Works 23 23

Experiments 32 service contracs from 5 different providers. 3 multi-constraint queries with increasing complexity. An ideal service contract rank is obtained as an agreement of a pool of experts. A modified version of the normalized discounted cumulative gain (NDCG) measure is adopted to assess the effectiveness of the proposed approach. Given a ranked result set S r and an ideal rank S i, the NDCG is evaluated as follows: NDCG( S, k) r = DCG( S DCG( S r i, k), k) 24

Experiments The NDGV average values at different @-cuts were evaluated considering different conditions : CASE 1: without considering the user profile; CASE 2: only by considering information taken at registration time; CASE 3: only by considering user history; CASE 4: only by considering i information on punctual values; CASE 5: only by considering information on textual description; CASE 6: the proposed approach. 25 25

Contents Problem Definition Motivation, Background and Contributions The Semantic and IR based Approach Multi-constraint query formulation Filtering and query evaluation Experimental Results Conclusions and Future Works 26 26

Conclusions and Future Works We propose a novel approach to service contract selection based on: definition of multi-constraint queries on precise and flexible preferences both explicitly defined by the users and implicitly inferred from their contexts; filtering of service contracts according to user category affiliations; evaluation of multi-constraint queries using semantic and IR techniques. Experimental results show the effectiveness of the proposed approach. Future works deal with: Building of a large benchmark of real service contracts; Management of contractual terms (e.g., security, trust) that cannot be directly quantified. 27 27

Thank you! Questions?

Future Works (?) Building of a large benchmark of real service contracts. How IR techniques can be used on available service contract descriptions (e.g., ProgrammableWeb)? Focus on specific domains or specific contract types (e.g., data contracts) in order to define the knowledge-base and reduce possible contractual al terms and values. IR techniques to support (functional) service discovery Evaluation of descriptions, service category and tags. Design and development of the semantic+ir service contract selector. 29 29