Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Visualization and text mining of patentand non-patent data

Anton Heijs

Treparel Information SolutionsDelft, The Netherlands

http://www.treparel.com/

ICIC conference , Nice , France , 2008

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Outline

Introduction

Applications on patent and non-patent literature

Patent mining and visualization

Mining and visualization Medline

Conclusions and future development

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Introduction

Mining + visualization = visual analytics

I More data - less insightI Overview needed

Text mining approaches controlling over/under-fit

Rule based : difficult to control under fitNatural language processing : difficult to control over fitMachine learning : controls over-fit and under-fit

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Document clustering : finding patterns unsupervised

Unsupervised text mining : clustering

I Input is a vector representation of all documentsI Separate the data in its natural groups based on a

similarity metricI Output is a similarity matrix

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Document classification : finding patterns supervised

Supervised text mining : classification

I Input is a vector representation of all documentsI Determine a hyper plane which separates all the data

I Binary classification : classify to 1 classI Multi class classification : classify to multiple classesI Compound classification : combine multiple binary

classifiersI Output is a matrix with classification scores

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Text analytics

Text analytics is text mining and visualization

I Combined use of mining and visualization for analysisI Uses text mining to determine trends in the dataI Use visualization to make complex data understandable

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Applications on patent and non-patent literature

Text mining and visualization steps

I management and preprocessing dataI text mining (filtering the data in the pipeline)I mapping filtered data to geometryI rendering and interaction

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Patent analytics applications

I Patent search using classification and clustering algorithmsI Patent landscaping, spotting new opportunitiesI Patent ranking, utilization optimization and valuation

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Search by querie by class

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Tree map visualization of the patent classification

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Clustering of patents of 7 classes in Optical Recording

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Clustering of large patent sets

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Building a classifier

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Testing a classifier

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Visualization of the classifier performance

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Patent analytics application combining mining and visualization

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Correlation between patents

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Patents over time

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Patents visualization hierarchically, supervised, unsupervised andover time

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Text mining and visualizations applications on Medline

I Search for facts and relationships (using classification andclustering)

I Disambiguation (using classification)I Named entity recognition (using classification)I Concept and relationships discoveryI Knowledge discovery by integrating text mining and

visualization withI Data mining : mining table data from experimentsI Image mining : micro array analysisI Graph mining : path way analysis

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Mining and visualization Medline

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Search Medline by querie

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Classification of Medline documents to multiple topics

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Visualization of trends in Medline documents over time

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Conclusions

I Combination of text mining and visualization algorithmsenables more application solutions

I Multiple application solutions are a combination of a smallset of text mining algorithms

I Multiple application solutions require multiple coupled viewvisualizations to provide complete

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Future development

I Patent analytics will include patent ranking and valuationI Text mining will integrate with data mining , for instance for

patent portfolio managementI Stand alone applications and applications with work flow

support will advance

Visualization and text mining Treparel

Introduction Applications Patent Analytics Medline Analytics Conclusions and future development

Trends Patterns Relationships

http://www.treparel.com/

Enabling you to learn more and see more

Visualization and text mining Treparel