New Technologies Supporting Technical Intelligence Anthony Trippe 221

New Technologies Supporting Technical Intelligence Anthony Trippe, 221 st ACS National Meeting

Aurigin Systems Inc. Aurigin Consulting Practice Director IP Consulting Services

Introduction w What is Technical Intelligence – Definitions – How Does it Fit with the Company’s Business Strategy • The Intelligence Cycle • Actionable Intelligence – What is it Not

Introduction (Cont. ) w Gatekeeper Approach to TI – The Intelligence Cycle w Ad-Hoc Team Approach to TI – The Intelligence Cycle

Introduction (Cont. ) w Computer Assisted TI – Data Mining – Text Mining w Available Methods – – Concept Clustering Self Organized Maps (SOMs) Neural Networks Decision Trees

What Is Technical Intelligence? w Definitions: – A tool to assist with long term strategic technical planning – Work processes for helping technical decision makers make smarter decisions faster – An analytical process that transforms disaggregated technological information into relevant strategic knowledge about your competitor’s technical position, size of efforts and trends

What Is Technical Intelligence? w How Does it Fit with the Company’s Business Strategy – Provides foresight into strategic activities • • • Entering new business areas Acquiring new technologies Evaluating competitor’s business moves Project guidance Developing partnerships

What Is Technical Intelligence? w Actionable Intelligence – Intelligence Cycle • • Define needs and prepare a plan Collect source materials Analyze the results Impact the business – Information when analyzed becomes intelligence – Intelligence directed towards a business decision becomes actionable – Must be used by the decision maker

What Is Technical Intelligence? w What is it Not? – – – For Patentability For Validity For Freedom to Practice Not about information its about intelligence It is about trends and forecasting not about focused and specific information retrieval

Gatekeeper Networks and The Intelligence Cycle w Define Needs and Prepare a Plan – Gatekeepers tend to be an expert in a specific area and typically only work in that area – TI is a part time job and involvement is often reactive – Tend to approach each problem the same way (hammer and nail approach) and while excited and interested in subject may not have time to stay current with new intelligence methods

Gatekeeper Networks and The Intelligence Cycle w Collect Source Materials – – Limited conference attendance Personal journal reading Personal networking Heavy reliance on the “grapevine”

Gatekeeper Networks and The Intelligence Cycle w Analyze the Results – Manual Mapping • Involves reading each document one at a time • Information is collected by using: – Spreadsheets – Word Processor tables – Flow charts – Butcher paper and sticky notes • Difficult to see hidden trends in large data sets • Does not scale well

Gatekeeper Networks and The Intelligence Cycle w Impact the Business – Delivers message using: • Handmade charts and graphs • Memos • Attendance at internal meetings – – Knowledge is power Potential silo creation NIH Potentially limited to specific projects

Ad-hoc Team Approach to TI w Define Needs and Prepare a Plan – Each project is done on a case by case basis using a team approach involving subject matter experts – TI Facilitators can communicate in a technically proficient manner and are trained in the field of TI with frequent updates – TI people are often employed full-time in conducting TI – Provides directed, actionable intelligence to the specific business need – The Need Drives the Question

Ad-hoc Team Approach to TI w Collect Source Materials – – – Size doesn’t matter Any available electronic source is fair game Print resources can be scanned in Internal and external data Also use human intelligence The Question Drives the Data

Ad-hoc Team Approach to TI w Analyze the Data – The Data Drives the Tool – Computer Generated Maps Can: • Group similar documents together • Build landscapes based on semantic concepts • Discover trends and do statistical analysis – Mining Activities • Data • Text – Does not replace reading the source materials

Ad-hoc Team Approach to TI w Impact the Business – Delivers message using: • Specific, focused charts, graphs and presentations • Detailed visualizations • Buy-in from subject matter experts – – Focused on business need Knowledge is shared Collective effort of many experts TI team is a corporate resource

Computer Assisted TI w Data Mining – Relies on fielded (structured) data and exact string matches – Involves numerically based statistical analysis – Allows for temporal analysis – Clustering based on coding – Involves co-occurancy matrixes • Examination of patent subject matter by Assignee

Co-code Clustering

Co-Occurancy Matrix

Graphical Representation

Computer Assisted TI w Text Mining – Relies on unstructured or semi-structured data – Term extraction takes place based on semantic based AI algorithms – Documents containing similar concepts can be organized together (Classification) – Documents containing overlapping concepts can be placed together geographically (Clustering)

Text Mining w Term Extraction Linguistic Preprocessing Reader Term Generation Tokens Candidate Generation Part of Speech Stemming Combination of Candidates Linguistic Patterns & Association Metrics Term Filtering Information Retrieval Metrics TFIDF

Text Mining w Information Extraction Term Extraction Co-Reference Named Entity Recognition Domain Knowledge Taxonomies

Available Methods w Concept Clustering – A form of SOM – Uses: • Term extraction • TFIDF • Bootstrapping and generation of vectors based on shared concepts – Topographical representation

Themescape

Available Methods w Self Organizing Maps (SOMs) – WEBSOM a method for automatically organizing collections of text documents and for preparing visual maps of them to facilitate the mining and retrieval of information – Details on SOM algorithm can be found at: http: //www. cis. hut. fi/research/som. shtml

WEBSOM

Available Methods w Neural Networks – Started as model of biological neural networks in the brain – Start with a training set – Use a second known set to measure difference between guess and known result – Computer makes adjustment, guesses again – Iterative process until within tolerance – Results visualized with standard methods (SOM, et…)

Available Methods w Decision Trees – Represents a set of rules – Training set identifies rules based on defined results and corresponding trends – Can be used on new data to make business decisions – Also called expert systems

Decision Tree