Smart Contract Risk Identification with AI
Download as PPTX, PDF1 like370 views
The document discusses the integral role of AI in contract risk identification and analytics, emphasizing the need to simplify data science for better user adoption. It details the data preparation and model-building processes for training AI on legal texts, highlighting categories such as confidentiality and payment terms. The focus is on combining human expertise with AI advancements to enhance contract review efficiency.
Smart Contract Risk Identification with AI
- 1. 1
- 2. 2 INTRODUCTION Ganes Kesari Co-founder & Head of Analytics “Simplify Data Science for all” 100+ Clients Insights as Stories Help apply & adopt Analytics Our data science platform, Gramex is now open- sourced!
- 3. 3 Smart Contract Risk Identification with AI
- 4. 4 AI VS LAWYERS: THE ULTIMATE SHOWDOWN https://hackernoon.com/20-top-lawyers-were-beaten-by-legal-ai-here-are-their-surprising-responses-5dafdf25554dhttps://www.lawgeex.com/resources/AIvsLawyer/
- 5. 5 CHALLENGES WITH CONTRACT REVIEW Voluminous Deliberate Verbosity Tedious Error-prone Lower cognition
- 6. 6 HOW CAN TECHNOLOGY HELP? Images: Gmail; Grammarly; Skype
- 7. 7 HOW DO THESE WORK? “Programs that solve the problem” “Programs that learn to solve the problem” vs Machine Learning
- 8. 8 HOW DO MACHINES LEARN? New Input Desired Outcome Machine learning how to do the job Known Input Known Outcome …210.5AAPL 207.4, 207.1,$204.5, 205.2,…
- 9. 9 ANALYTICS IS MORE THAN MAGIC WITH NUMBERS ”What a nice sunny …day? Icons: Watercolor vector by starline on freepik; flaticon on freepik; all-free-download.com
- 10. 10 WHY DEEP LEARNING? 10 Input Output Identify features to teach model Traditional Machine Learning Deep Learning Person Name Input Output Model automatically identifies features to learn Person Name https://www.cs.toronto.edu/~ranzato/publications/taigman_cvpr14.pdf
- 11. 11 NOT VERY DIFFERENT FROM HOW WE LEARN Training ..Versatile Detection!
- 12. 12 AI AROUND US - READING TEXT
- 13. 13 AI AROUND US – TRANSLATING TEXT Google Translate
- 14. 14 AI AROUND US – GENERATING TEXT OpenAI: https://openai.com/blog/better-language-models/
- 15. 15 OUR WORK ON CONTRACT RISK IDENTIFICATION ✔️ Categorize ✔️ Assess Risk ✔️ Simplify Reviews
- 16. 16 PREPARING DATA TO TRAIN THE AI 100+ Documents 6000+ Clauses Icon by Smashicons on Freepik Categories: Agreement Liability-Indemnity Confidentiality Others Dispute Resolution Payment Terms IP Rights Scope Terms And Conditions Sub Contract Warranty Data:
- 17. 17 Term. This Agreement shall be in force for a period of two (2) years from the Effective Date. This Agreement may be terminated by either Party at any time upon thirty (30) days written notice to the other Party. Provided however that the termination of this Agreement shall not relieve any Party of its obligations with respect to Confidential Information disclosed under this Agreement for a period of five (5) years after the expiry/termination of this Agreement. PREPARING DATA TO TRAIN THE AI Categories: Agreement Liability-Indemnity Confidentiality Others Dispute Resolution Payment Terms IP Rights Scope ✔️ Terms And Conditions Sub Contract Warranty Criticality: Non-Critical ✔️ Critical
- 18. 18 BUILDING THE MODEL Data Preparation Gathering Structuring Labeling Modeling ULM Fit Language model Built a Classifier – Criticality & Clause type Transfer Learning Pretrain on WikiText Train on Legal text User Workflow Interactive UI Human augmentation Start End
- 19. 19 LANGUAGE MODEL PRIMED ON OUR CONTRACTS Gramener Technology will be…. 1. Gramener Technology will be… the sole and exclusive owner of Gramener ’s Confidential Information and the Project Plan is for the sole purpose of exploring the possibility of Data Analytics , Visualization and Consulting Services . xxbos This Master Services Agreement ( “ Agreement ” ) is made on 16th February , 2015 ( “ Effective Date ” ) xxbos The Software Services and Maintenance Services Guidelines will be provided by the Client. 2. Gramener Technology will be… a company incorporated under the provisions of the Companies Act , 1956 or having its registered office at Plot No.9 / 2 , 2nd Floor , Sy . No.64 , HUDA Techno Enclave , Phase – II , Madhapur , Hyderabad - 500081 , Telangana , India ( “ Gramener ” ) , ( " Vendor " ) and the place of supply to Client ( hereinafter referred to as " the Receiving Party " ).
- 20. 20 DEMO
- 21. 21 TAKEAWAYS Acquire right Data Structure & Clean Data Plan for labelling Factor business dynamics Augment with human inputs Build into workflow Sensitize on accuracy Plan for refresh
- 22. 22 HUMAN + AI “ Machines will partner and cooperate with humans rather than become mankind’s biggest enemy. - Jack Ma, Founder of Alibaba
- 23. 23 @kesaritweetsgramener.com @kesari Presentation deck available at gkesari.com/LawAI Thank You!
Editor's Notes
- #22: Acquire data – Water, water around us, but not a drop to drink! Data structure – archived as scanned documents! Label your own data – manual task that people hate to do! Look out for practical challenges – fuzzy classifications Augment with human intelligence – AI to speak up when unsure! Build it into the user’s natural workflow – key for adoption! Sensitize users on low accuracy – It’s not a SAT score Plan for model refresh – how do you keep it alive?