TETRALOGIE

1
TETRALOGIE

• Competitive Intelligence Software
• http//atlas.irit.fr
• B. Dousset (I.R.I.T.) dousset_at_irit.fr

2
I.Competitive Intelligence
3
Environment Observation and Analysis

• Public organizations
• Governments policies
• Administration
• Research
• Industry
• Medium and big size, ...
• Research and development team
• Management
• Education

4
Evolution Analysis of

• Activities
• Scientific,
• Technical,
• Economical, ...
• Markets
• Cooperations
• Participations
• Innovations
• Technologic transfers
• Employment trends

• Research teams
• Terminology
• Databases, Patents
• News groups
• Web
• Publications
• Press
• International legislation
• Industrial property

5
II. Context ofDevelopment
6
Partners of Bibliometry Station "ATLAS"

• Defence Ministry
• CEDOCAR
• SGDN
• DRET
• DGA
• Regional Administration
• ADIT
• SANOFI

• CNRS
• IRIT (Toulouse)
• INIST (Nancy)
• CERESI (Meudon)
• CRRM (Marseille)
• INRA
• INSERM
• Universities

7
Cooperations

• Ministries
• Universities
• Libraries
• Research centers
• Public organisations
• Private Industries
• Documentation experts
• Journalists,

8
Trainings (five days)

• Defence Ministry
• Home Office
• INSERM, INRA, CNRS
• Sanofi, P. Fabre
• EDF, GDF, CEA
• Aérospatiale
• Total, Elf, IMP
• Nestlé, L'Oréal, Bourjois
• Psa, France télécom

• Ecole des mines
• LAAS, INIST, ADIT
• CIRAD, UCIST
• PerioMip
• URFIST
• Delta-Veille
• Avant-Veille
• Cybil, Desco
• Virtualis, Corum

9
Implemented Workstations I

• Defence Ministry (Paris/3 SUN et PC)
• Home Office (Paris/SUN)
• Treasury (Paris/SUN)
• IRIT/CNRS (Toulouse/SUN) server
• LAAS/CNRS (Toulouse/SUN) server
• INRA (Toulouse and Paris/SUN) server
• INSERM (Strasbourg and Paris/SUN) server
• CEA (Grenoble/HP)
• DGA (Paris/PC)
• CEDOCAR (Paris/SUN)

10
II

• INIST/CNRS (Nancy/SUN)
• CRRM/CNRS (Marseille/IBM et PC)
• LVI (Marseille/PC) server
• CIRAD-UCIST (Montpellier/HP)
• CICT (Toulouse/HP) permanent training
• ADIT (Strasbourg/SUN et PC)
• Polytechnic school (Barcelona Spain/SUN)
• University of Columbia (Bogota/PC)
• University of Mexico (Mexico/SUN) server
• University of Marne la vallée (Paris/SUN)
• Mine school (Albi/SUN)
• Medical university (Marseille/BULL)

11
III

• NESTLE (Lausanne Switzerland/PC)
• BOURJOIS (Paris/PC)
• TOTALFINAELF (Paris Pau/SUN)
• IMP (Mexico city Mexico./SUN) server
• GFI Benelux (Luxemburg/SUN ) server
• CARMEUSE (Seilles Belgium/PC)
• PSA (Paris/Sun)
• France Télécom (Paris/Sun)
• Thales (Paris/PC)
• IALE Tecnologia (Barcelona Spain/PC)
• LEIA (Victoria Spain/PC)
• CTBA (Paris/PC)

12
Tétralogie at the UPS/IRIT

• A "Tétralogie" server and Web
• Development two workstations (IRIT)
• A workstation devoted to analysis (Library)
• Personal Computers used as
• Bases requesting station
• X terminal
• Office automation tools
• Linux "Tétralogie" server
• PC with modem for remote working
• Portable PC for demonstration

13
(No Transcript)
14
III. Components of Tétralogie Software
15
Corpus Manipulation

• Applications management
• Bases knowledge inheritance
• Bases structures descriptors
• Descriptors and tools management
• Data correction (format)
• Searching for vocabulary
• Making filters and dictionnaries
• Making tables of co-occurrency, contingency
• Making binary tables, 3D tables

16
Specific 3D Spreadsheet

• Table cleaning
• New table extraction
• Relinearisation
• Normalisation
• Comparisons
• Map of table
• Panoramic Zoom
• fish eye vue

• Sorting out
• Consistance
• Alphabetical order
• Absolute block
• Relative block
• Connexity
• Histograms
• Pre classifications
• Import/Export

17
Analysis Methods

• Principal Components Analysis (PCA or ACP)
• Correspondances Factorial Analysis (CFA or AFC)
• Visualisation of factorial mapping (2D, 3D, 4D)
• Visualisation of quantitative links in 4D
• Classification methods
• CAH hierarchical and CPP partitional methods
• Study of evolution
• absolute evolution (trajectories)
• différential evolution (speed and acceleration)
• relative evolution (procustean rotations)

18
IV. Using Modes
19
Monoposte Aspect

• Multi tasking (parallelism process)
• Multi windowing (comparison of results)
• IPC (Inter Process Communication)
• Collaboration between analysis methods
• Descendant analysis
• Progressive elimination of typical cases
• Permanent control of quality
• Macros processes
• Incremental analysis (in time)
• Links with PC (Word/Excel)

20
Network Aspect I

• Direct Access to InterNet data bases
• Wais, 3W, News Groups,
• Bibliographic servers, Patents, ...
• Connexions
• from X terminals
• from UNIX workstations
• from PC (Xwin, ReflectionX, PcXWare, Exceed, ...)
• from Macintosh (MacX, ...)
• Free use of the IRIT "Tétralogie" server
• Multi connexions (tele working)

X emulators
21
Network Aspect II

• Records of remote analysis strategies
• Server of analysis already made by experts
• Shared analysis
• Shared corpus
• Distant expert interventions
• Distant analysis made in real time
• On line sofware installation and tests
• On line software updating
• On line demonstrations

22
Tétralogie Use

• On our workstation or PC
• (SUN, HP, IBM, Silicon, Bull, PC-Linux, ...)
• Through a connection to "Tétralogie" server
• (IRIT, CICT, BU, InterNet or IntraNet, ...)
• Analysis made by a "Tétralogie" specialists
• (IRIT, BU, ADIT, Delta Veille, Avant Veille)
• By consulting the already existing analysis
• (IRIT SUN, IRIT PC, IRIT HP, )
• By connecting to the site
• http//atlas.irit.fr

23
A Coherent Training Offer

• Course I Initiation training (three days)
• Course II Advanced training (two days)
• Thematic session for user support
• Analysis starting session
• Advanced concept session
• Training to collegiate working on InterNet
• University training (DESS, IUP, School, ...)
• Conferences and exhibitions
• On line documentations and publications

24
V. "Tétralogie" Tools
25
"Tétralogie" V6.0

• One X-Window
• Parent directory
• Access path
• Analysis management
• New application
• Scrolling menu
• Applications list
• Software exit

26
New Application Function

• Name of application
• Select
• bases
• formats
• Inheritance
• structure descriptor
• format corrector
• specific tools
• dictionnaries
• filters, ...

27
Analysis Management

• Treatments ...
• Conversion DOS format to UNIX format
• Conversion Html format to UNIX format
• Application compression gzip factor / 20
• Application uncompression gunzip factor 20 X
• Cleaning corpus
• Correction of format

28
Corpus Preparation

• Treatments
• Countings
• Crossing tables
• Presence/Absence
• Co-occurrence
• Frequency
• Evolution study
• Multi Co-occurrence
• Synonymy
• Multi-Terms

29
Analysed Corpus Format

• Fields
• Sub-fields
• Separators
• Items
• words
• names
• dates
• codes
• Free text
• Keywords

30
Format Descriptor
Name of field Abbreviation Field in data
base Mask Separators
31
Counting Process

• Fields choice
• Cutting
• frequency or
• Filters
• positive
• negative
• Synonyms
• Files creation
• .ind, .indA, .indF
• Load curve

32
Synonymy Process

• Several strategies
• Authors
• Keywords
• Address
• Making file .Syn
• Auto proposition
• Manual validation
• Fusion
• Recounting
• Good frequencies

33
File Manager

• On UNIX
• Access to
• Corpus
• Filters
• Dictionaries
• Synonyms
• Cutting
• Results
• Reports
• Shell UNIX, ...

34
Co-occurrence Process

• Making 2D tables TI-TI, TI-PA, OR-PA, ...

35
Functions Menu

• 2D and 3D Spreadsheet
• Methods
• ACP, AFC
• CAH, CPP
• Linking, Procuste
• Visualisations
• 2D, 3D, 4D
• Animations
• 3D, 4D
• Editions and Utilities

36
VI. Tables Manipulations
37
Spreadsheets

• Tables management
• Capture options
• Sorting functions
• Import/Export
• Transformations
• Editions
• Zoom out
• Interactivity
• Histograms
• Selections
• Cutting, ...

38
Lines or Columns Histograms

• Simple
• Line CNRS function of the time divided in
  homogeneous periods
• Comparative
• CNES compared to LAAS according to time

39
Mathematic Transformations

• Linearisation
• Percentage
• Reduction
• Absolute value
• Division per
• Marginal of lines
• Marginal of columns
• Square root of marginals
• Square root of diagonals
• Global or local mode

40
Sorting by Consistance

• Sorting lines and columns
• Sorting one line
• Sorting one column

• Searching for
• Teams
• Leaders
• Typologies
• Emergence
• Terms
• Concepts
• Journals
• Autors
• Countries
• Links

41
Lines and rows connections
42
2D Zoom
43
3D Zoom
44
Connexity Sorting

• Non-connex table

• Connex classes

Zoom out
45
Small Links Removal

• Partial disconnection
• Connexity sorting
• Detection of
• Sub class
• Classes number
• Links between classes
• Comparison with
• The biggest eigen values
• AFC in 4D vue
• Classifications

46
Pseudo Diagonal Blocks Sorting

• Used when the table is connex
• Detection of clusters (homogeneous groups)
• Put then along the pseudo diagonal
• Zoom-out of table
• allows to show clusters
• enables to navigate along diagonal
• Possible extraction of clusters
• Descendant analysis with AFC
• Comparison with classifications
• Detailed content of lines and columns

47
Clustering of Blocks Sorting
48
Co-citations of Patents
Patents with same referencies
Boeing Patents (Uspto 76-99)
49
Pre-classification

• Enables to characterize the individus
• Classes number in the ad hoc column
• Makes coloration
• in 3D vue
• in 4D vue
• Sorts consistance
• Extracts classes
• Compares with
• AFC
• Classifications

50
Multi-terms Detection

• Choice of an updated dictionary
• Removal of empty words
• Synonymy
• Treatments
• Syntactic
• Morphologic
• Statistic (threshold)
• Rewriting
• In a new field
• New indexation

51
Emergent Concepts
Diagonal cluster
52
VII. Graph Analysis Method
53
Complex Relations Visualization
With Low links
54
Simplification of graph
Without low links
55
Clustering and connectors
Links between teams
56
Clusters Zooms
Team B
Team A
57
Case of asymmetrical matrix
Outstanding links
58
Sorting of items
Cut of class
Outstanding links
59
Extremal partial tree
60
Reduced tree
61
Detail of Item
62
VIII. Multidimensional Analysis Method
63
Principal Component Analysis

• Application to factual data (frequencies)
• Representation lines in the columns space
• Search on the main axis of this one
• Better visualisation in 2D, 3D, 4D vues
• Representation of correlation circle of columns
• Comparison of the two vues to find relations
• Exhibition with selections and links
• Descendant analysis of other factorial axis

64
Reduced Principal Component Analysis

• Same process as in ACP
• Data are centred
• Data are reduced
• The size effect is suppressed
• Eigen values are lower
• The descendant analysis is more advanced
• Possibility of variables classification

65
2D and 3D Visualisations

• 2D Visualisation

• 3D Visualisation

66
4 D Visualisations
67
4D Functions

• Exportations
• Rotations
• Axis sliding
• Edition

• Imports
• Rotations
• Axis sliding
• Classification

• Zoom in/out
• 4D Rotations

68
Correlation Circle (or 4 Sphere)
69
Correspondances Factorial Analysis

• Profiles ACP
• Qualitative data
• Binary data
• Contingence table
• Only one map
• Lines (individuals)
• Columns (variables)
• Typology analysis
• Proximity analysis

• Quality of analysis
• Associated parameters
• Eigen values
• Eigen vectors
• Inerties
• Cumulated values
• Correlations
• Densities
• Relative contributions
• Absolute contributions

70
Quality Observatory of Analysis
71
4D Visualisation of AFC
Median field
Field with team
Median field
Median field
Field with team
72
4D Vues in Concordance(Rotation Export)

• Variables (columns)

• Individuals and Variables

73
Multiple Selection Icon Class
Year 97

• Keeping
• Selection
• Rest
• Extraction
• Items
• Documents
• Coloration
• Selection

Emergent Items
74
Links Visualisation

• Links beetwen
• Individuals
• Variables
• Individuals Variables
• Identification of
• concerned Actors
• Links strength
• Nodes
• Isolated elements
• Searching
• By 4D rotations
• Axis sliding

75
IX. Automatic Classification Methods
76
Hierarchical Ascending Classification

• Decomposition by overlapping
• Different distance types
• Euclidian
• Maximum
• Minimum
• Medium, ...
• 2D Tree representation
• Exportation to 4D vue or spreadscheet

77
Class Exportation in 4D vue
Superposition of classes
Cutting line
Principal plane is not good choice
78
Classification Navigation
LAAS (Toulouse) Thematical Classes
Team
Cutting line
79
Partition Classification

• Non hierarchical method
• Initialisation of classes number
• Maximum inter class distance
• Minimum intra class distance
• Interactive process
• Two methods
• Mobile Centers (one center per class)
• Dynamic clouds (Diday)
• 3D visualisation

80
Classes Choice and Results
81
Classification by Expert System

• Classification explanation
• Common points in each class
• Objective choice of similarity criterion
• Decisional model
• Comparison between
• ACP method
• AFC method
• Sorting methods

82
X Evolution Analysis
83
The "Time" Variable

• First method (2D)
• Using items of date as variables (Columns)
• Defining homogeneous periods (ex 94-96, 97-99)
• Working in relative values ( / Periode)
• Sorting by consistance in the last periode
• Second method (3D)
• Making one table per period (3 or 4 tables)
• Analysing with AFC
• one period (others are projected)
• all periods (medium analysis)
• sum of periods (idem static case)

84
Multi Cooccurrence Table
Stable Collaboration
Loss collaboration
New collaboration
85
Trajectories exploratory
Team change
n of périod
2 years collaboration of team BEROT  co
86
In line Items Trajectories
Trajectory of BABAGUCHI, N.
Trajectory of NAKAMURA A.
Expert Concentration
Trajectory of KOBAYACHI S.
87
Procustean Rotations
Abnormal evolution
Study of evolution quality
Normal evolution
88
Speed and Acceleration

• Subtraction of tables
• Differential values generation
• Speed vector
• T1, T2, T3, T4 gt T2-T1, T3-T2, T4-T3
• Acceleration vector
• T2-T1, T3-T2, T4-T3 gt
• (T3-T2)-(T2-T1) et (T4-T3)-(T3-T2)
• soit T3-2.T2T1 et T4-2.T3T2
• Tendance analysis
• Prospective method

89
IX.Geo-strategy
90
Interactiv Geographic MapsInternational
Collaboration (INRA)
Choice of representation
91
Choice of Geographic Area

Choice

92
Choice of Color Scale

Linear scale


Choice

Scale

93
Choice of Weighting

Absolute Values


Population
Weighting Choice

National Product
Area


94
Classification Importation
Line cut
95
Segmentation
Segmentation
96
XI. Actual Possibilities and Perspectives
97
Specific studies

• State of the art in a research field
• Follow up of the evolution of a field
• technical, economical, juridical, ...
• Position study of
• organization, firm, research team
• Strategical component evaluation
• Site Web analysis (connexion)
• Research evaluation
• Experts detection, ...

98
Cooperation with Experts

• Definition of the best analysis scheme for each
• type of base
• patent format
• datawarehouse structure
• Elaboration of specific tools
• filters, dictionaries, synonyms, ...
• Multi bases analysis
• "Tétralogie" experts qualification
• Delivery service assent in CI

99
Cooperation with Users

• Implementation of more powerful servers
• at IRIT (analysis and development)
• at the university library (CD exploitation)
• for trained users
• Sharing tools or servers
• On line documentation
• Automatic updating software
• "Tétralogie" users club
• Telephonic support
• Specific developments

100
TétraWeb Server

• Users forum on Web
• Dates, location and content of permanent training
• Access mode to servers
• Users list
• Expert list and coordonnates for each field
• List of exhibitions on competitive intelligence
• New "Tétralogie" tools and using notice
• Scientific publications of our team
• More analysis examples with "Tétralogie"

101
http//atlas.irit.fr
102
Software Evolution

• Extension of the "ATLAS" norm with our partners
• Improvement of existing methods
• Addition of new analysis methods
• Extension to language processing
• Full text, abstract, html format, patents, press,
  ...
• Software use on the Web
• Push methods
• Automatic analysis (Macros)
• Multi language version (7.0) english, spanish
  !!!