ai-fingerprints: finding the intrinsic patterns in language

24
© ai-one inc. 2012 biologically inspired intelligence ai- © ai-one inc. 2012 ai-Fingerprints: Finding the intrinsic patterns in language

Upload: ai-one

Post on 22-Jun-2015

2.810 views

Category:

Technology


1 download

DESCRIPTION

How to generate a knowledge representation model for unstructured text.

TRANSCRIPT

Page 1: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012

biologically inspired intelligence

ai-one™© ai-one inc. 2012

ai-Fingerprints: Finding theintrinsic patterns in language

Page 2: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012

Biologically Inspired Intelligence

creativitylogic

© ai-one inc. 2012

Page 3: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012

Detects the meaning of a document by identifying the most semantically important words in the text.

Automatically generates a graph representation of knowledge.

Think: Feature detection of vertices and edges (keywords and associations) in any document, in any language.

Works without human intervention or models.

Make sense of any* text

© ai-one inc. 2012

* as long as there is sense in the unstructured sentences…

Page 4: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012

I0I00I0II0I0III0I0I00II0II0I00II0I0II0I000II0I0II0I00II0I00II0I0II0I000I0II0I0II0I0I0II0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II00II0I0I00II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0I0I0I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00II0I0I00II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II

ai-one

Page 5: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012

I0I00I0II0I0III0I0I00II0II0I00II0I0II0I000II0I0II0I00II0I00II0I0II0I000I0II0I0II0I0I0II0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II00II0I0I00II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0I0I0I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00II0I0I00II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II

Page 6: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012

ANY Language!

Page 7: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012

ANY Language!

Page 8: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012

prolitterisdorerchristianvaldaandreasblick20020129seite5nummer23swissairboss12miofür5jahreimvorauskassiertderdruckwächstcortiindeckungvonchristiandorerundandreasvaldazürichesistvölligunüblichdassderlohnfürfünfjahreimvorausbezahltwirdsagenrenommierteheadhunterfdppräsidentgeroldbührer53verlangtjetztschonungslostransparenznochimmersagenmariocortiundseinefreundeausdemaltenverwaltungsratnocommentcortiweiltegesterninpolenundwolltenochimmernichtsagenoberaufeinenteilseines5jahreslohnesverzichtenwirdinschweigenhüllensichauchdieehemaligenverwaltungsrätedieimmärz2001denvergoldetencortivertragaufgesetzthattenindersalärkommissionsassdamalszementkönigthomasschmidheiny56herrschmidheinymöchtesichnichtäussernsolangedieuntersuchungdessachwaltersläuftlässterausrichtengleichtöntesbeimzweitenkommissionsmitgliedgaudenzstaehelinichhabeverständnisfürdasinteresseaberesliegtanderswissairzuinformierennichtanmirauchdieübrigenmitgliedervonvrenispoerrybislukasmühlemannverkriechensichdabeiwarenauchsiebestensimbildsolcheverträgewurdenimmerimplenumbesprochensagteinehemaligesvrmitglieddochcreditsuissechefmühlemann51bequemtsichgenausowenigzueineroffeneninformationwiebankierbénédicthentsch53amtelefonsagterentnervtichhabedasrechtnichtszusagenwarumwollensienichtssagenherrhentschfragensienichtweiterichgebekeinenkommentarschönentagaufwiedersehenaufgehängtjetztregtsichpolitischerwiderstandwennderbetragohneauflagenüberzieleausbezahltwurdedannistdasinakzeptabelsagtfdppräsidentgeroldbührerüberfdpmitgliedcortijetztbrauchtesschnelltransparenzdenkopfschütteltauchcvppräsidentphilippstähelin57ichhabemühemiteinemsohohenlohndasführtzuriesigeneinkommensunterschiedenimvolkdasdarfnichtseinerhaltezwarvielvonleistungslohndochdenkannmannichtimvorausbekommenvorauszahlungensindunüblichsagtheadhuntersandrovgianellaundfredyislervonspencerstuartichhabenochnievoneinemsolchenfallgehörtaberniemandwolltedenswissairjobdiesesversprechenwarwohleinlockvogeldamitcortiseinensicherennestlépostenaufgabmariocortilukasmühlemannthomasschmidheinybénédicthentschkonzernsanierermussvorgerichtzürichdervorkassevertragcortisistkeinepremiereimfallderkonkursitenbiberholdingliesssichkonzernsaniererchristianspeiserbildseinenjobmit28millionenfrankengarantielohnvergoldenermusssichjetztvorgerichtverantwortenwiecorti

ANY Language!

Page 9: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012

I0I00I0II0I0III0I0I00II0II0I00II0I0II0I000II0I0II0I00II0I00II0I0II0I000I0II0I0II0I0I0II0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II00II0I0I00II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0I0I0I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00II0I0I00II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I00I

ANY Language!

Page 10: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012

I0I00I0II0I0III0I0I00II0II0I00II0I0II0I000II0I0II0I00II0I00II0I0II0I000I0II0I0II0I0I0II0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II00II0I0I00II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0I0I0I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00II0I0I00II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I00I

ANY Language!

Page 11: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012

I0I00I0II0I0III0I0I00II0II0I00II0I0II0I000II0I0II0I00II0I00II0I0II0I000I0II0I0II0I0I0II0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II00II0I0I00II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0I0I0I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00II0I0I00II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I000II00II0II0I00II0I0I0I00III000I0I0III0II00I0II0I00I0I0I0I0I00I0IIII0I0I0I0II00I0II0I00I

Light Weight Ontology (LWO)!

Page 12: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012© ai-one inc. 2010

obesity

lifestyle

nutrition

physical exercise

stress

smoking

0.9

0.90.60.75

0.9

Our technology detects the strength of all associations between all words (n:n).

The LWO captures the intrinsic semantic value of every word, sentence, etc.

ai-one is self-learning, self-organizing and incrementally self-updating.

Lightweight Ontology (LWO)Association Strength

Page 13: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012

Sender

Subjective interpretation & understanding

subjective meaning and semantic

Every human interprets semantics differently because of experience, intellectual level, cognitive biases, etc.

Be careful: The semantic trap!

Receiver

ai-one has the neutral observer position to detect the inherent meaning

ai-one shows the LWOregardless if right or wrong

Page 14: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012

From LWO to one ai-fingerprintConvert the LWO-pattern in a simple visualization with graphs that show, store & match fingerprints

Page 15: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012

• jobs -> c.v.• c.v. -> jobs

c.v.

job

job

job

c.v.

c.v.

job

job

• data in forms -> free text• free text -> forms

• Dynamic• Intelligent

• similar text • similar meaning

Use case: ai-fingerprint to match job posting with resumes for HR

Page 16: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012

The fingerprint is defined by the the most important words of a text corpus – called keywords. Keywords words are classified in importance according to a value determined by ai-one.

Keywords can be used to search with Yahoo, Google, Wolfram Alpha, Bing and any other search approach. Since the keywords are the optimal search words, users will get the best answer by traditional search engine.

Spicing with the ai-fingerprint semantic for better results.

ai-fingerprint as bridge to traditional approaches

Page 17: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012

We solve the problem of identifying multiple themes in a single document. Very often, there are multiple items, issues or themes on the same page.

For example in a news magazine, there are short article about different subjects on the same page. This is very challenging for traditional semantic text analysis processes.

ai-one’s LWO automatically segments topics into related clusters.

ai-fingerprint solves an other problem in traditional text analytics

Page 18: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012© ai-one inc. 2010

ai-Fingerprint Typical ai-fingerprint of a text corpus very focused on one issue/theme.

Page 19: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012© ai-one inc. 2010

ai-Fingerprint Typical ai-fingerprint of a text corpus where tree different issues/themes are on the same page

11

22

3

3

Page 20: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012© ai-one inc. 2010

ai-Fingerprint Typical ai-fingerprint of a “nonsense”-text corpus where no real sentences are giving an story, randomly sentences

Page 21: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012© ai-one inc. 2010

ai-Fingerprint Enables better decision making by providing a way to compare semantic content by using graphs.

Page 22: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012

… much better than any traditionally method!

Current linguistics and semantic solutions work only if they are feed accurate and detailed language models.

These models are expensive and specific to a single dialect.

Worse, they do NOT incrementally update or learn!

© ai-one inc. 2010

…ai-one –

Page 23: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012

… recognizing the content… understanding the meaning and

generalizing its application… deciding about its importance… knowing what to do with this

learned information

© ai-one inc. 2010

…ai-one – the next evolution in software… intelligent agents?

Page 24: ai-Fingerprints: Finding the intrinsic patterns in language

© ai-one inc. 2012

Thank You!

© ai-one inc. 2012

ai-one inc. 5711 La Jolla Blvd., Bird RockLa Jolla, CA 92037

[email protected]

ai-one ag Flughofstrasse 55, Zürich-Kloten8152 Glattbrugg

ai-one gmbh Koenigsallee 35a, Grunewald14193 Berlin