CSEG3060 · Research Methodology in Computer Science

Patent Information and Administration

Patent literature · IPC and CPC classification · the major databases and how to search them · and how the patent system is actually run

Roughly 70% of the world’s technical information is disclosed only in patent documents, and never published anywhere else.

Unit VI · Lecture 3 of 4 · 60 minutes · CO4 · B.Tech (CSE) Semester V

Dr. Mohsin Furkh Dar · School of Computer Science, UPES

→ / Space next · ← back · O overview · N speaker notes · F full screen

Learning outcomes · CO4

By the end of this session, you can…

01

Identify the components of a patent document and read a front page using INID codes

02

Explain the IPC hierarchy and decode a classification such as G06F 21/62

03

Distinguish CPC from IPC, and name the CPC codes covering your own research area

04

Navigate PATENTSCOPE, USPTO, Espacenet and IPAIRS, and choose the right one for a given task

05

Formulate a search combining classification codes, keywords and Boolean syntax

06

Describe the three tiers of patent administration, the prosecution workflow, and recent reforms

CSEG3060 · Unit VI · Lecture 3 — Patent Information, Databases and Administration2

Patent literature · 1 of 2

The largest technical corpus nobody reads

Patent literature is the entire corpus of published patent documents and related materials generated by national and regional offices. WIPO holds it to be the world’s most comprehensive source of technical information — with about 70 percent of all technical information disclosed only in patent documents and never published elsewhere.

Five components of a patent document

Bibliographic dataNumber, date, country, applicant, inventor, classification codes
AbstractConcise technical summary
DescriptionFull disclosure enabling a person skilled in the art to reproduce it
ClaimsThe legally enforceable statements defining scope
DrawingsDiagrams, flowcharts, schematics

Primary and secondary, and the kind codes

Primary — applications and granted patents published by the offices.
Secondary — search reports, examination opinions, gazettes, classifications, abstracting services.

A1 application published with a search report · A2 application published without one · B1 granted patent with no prior application publication · B2 granted patent where the application was already published

Utility models — shorter-term protection for incremental innovation, common in Germany and China. They are prior art too.

CSEG3060 · Unit VI · Lecture 3 — Patent Information, Databases and Administration3

Patent literature · 2 of 2

Why it matters to you, and how to read the front page

Prior-art search

Has the algorithm or architecture already been patented, before you file?

Freedom to operate

Which patents might your planned commercial product infringe?

Landscape studies

Mapping innovation trends in ML, blockchain, quantum computing, cybersecurity.

Citation analysis

Tracking how patented technologies build on earlier scientific work.

Licensing

Identifying patent holders open to a licensing arrangement.

The front page — INID codes under WIPO standard ST.36

(10) Document identification — patent number, kind code
(20) Application data — number, filing date

(30) Priority data — earlier applications under the Paris Convention
(40) Publication date

(50) Technical information — IPC/CPC classification, title, abstract
(60) References — related applications

(70) Identification of parties — applicant, inventor, assignee, agent

INID stands for Internationally agreed Numbers for the Identification of bibliographic Data. Knowing the numbers lets you read a patent from any office in any language — the field codes are constant even when the text is not.
CSEG3060 · Unit VI · Lecture 3 — Patent Information, Databases and Administration4

Classification · 1 of 4

The IPC — five levels, decoded

The International Patent Classification, established under the Strasbourg Agreement (1971) and administered by WIPO, is the international standard for classifying patents and utility models. It is revised annually to accommodate emerging technologies.

Five-level hierarchy

Section → Class → Subclass → Group → Subgroup

Worked example — G06F 21/62

G  —  Section: Physics

G06  —  Class: Computing; calculating; counting

G06F  —  Subclass: Electric digital data processing

G06F 21  —  Group: Security arrangements for protecting computers, components, programs or data against unauthorised activity

G06F 21/62  —  Subgroup: Protecting access to data via a platform, e.g. using keys or access control rules

The version number is appended to indicate the edition used — for example IPC 2025.01. Most computer science patents fall in Section G (Physics) and Section H (Electricity).

CSEG3060 · Unit VI · Lecture 3 — Patent Information, Databases and Administration5

Classification · 2 of 4

The eight sections — and the two that concern you

SectionTitleDomain
AHuman necessitiesAgriculture, food, health, personal items
BPerforming operations; transportingIndustrial processes, transport
CChemistry; metallurgyChemical engineering
DTextiles; paperTextile and paper industries
EFixed constructionsBuildings, mining
FMechanical engineeringEngines, pumps, lighting, heating
GPhysicsInstruments, computing, optics, nuclear engineering
HElectricityElectric power, electronics, communication
Almost everything a computer scientist patents sits in G06 (computing) or H04 (electric communication technique). Learning those two branches is a far better use of an afternoon than memorising all eight sections.
CSEG3060 · Unit VI · Lecture 3 — Patent Information, Databases and Administration6

Classification · 3 of 4

The CPC — finer, faster, and used by the two biggest offices

The Cooperative Patent Classification was jointly developed by the EPO and the USPTO and entered into force on 1 January 2013. It is built on the IPC backbone but extends it with finer subdivisions.

Four differences from the IPC

  • Depth — roughly 200,000 additional codes beyond the IPC’s ~70,000
  • Tagging — schemes for cross-cutting technologies
  • Update cadence — revised every three months, against the IPC’s annual cycle
  • Adoption — used by both the EPO and the USPTO, harmonising the two largest granting authorities

CPC codes for computer science

G06N 3/04Neural network architectures
G06N 3/08Learning methods for neural networks
H04L 9/00Cryptographic mechanisms
H04L 9/06Block cipher systems
G06F 16/00Information retrieval; database structures
G06T 7/00Image analysis
G10L 15/22Speech recognition procedures
G06Q 50/18Legal services — including IPR itself

Find the two or three codes covering your own project and note them now — they are the entry point for every search in the second half of this lecture.

CSEG3060 · Unit VI · Lecture 3 — Patent Information, Databases and Administration7

Classification · 4 of 4

Why you need both — and what else exists

WIPO and national offices

WIPO and many national offices — including IP India — still index primarily by IPC.

EPO and USPTO

Both use CPC extensively. If you search only by IPC on Espacenet you are discarding the finer granularity.

The practical rule

The additional granularity of CPC allows far more precise searches in cutting-edge technologies — which is exactly where a CS researcher works.

Other systems worth recognising

USPC
United States Patent Classification — used by the USPTO until 2015, when CPC fully replaced it. You will still meet it on older documents.

FI and F-term
Japanese systems offering very fine subdivisions — valuable for electronics and imaging prior art.

DE
The German national scheme used by the DPMA.

For a comprehensive search, researchers cross-reference multiple systems. A single-classification search is the equivalent of searching one database and declaring the field empty.
CSEG3060 · Unit VI · Lecture 3 — Patent Information, Databases and Administration8

Databases · 1 of 3

WIPO PATENTSCOPE and the USPTO tools

WIPO PATENTSCOPE

Covers: PCT international applications published 18 months after priority · national and regional collections from over 100 offices · non-patent literature cited in search reports.

Key features: Cross-lingual information retrieval (CLIR) searching several languages at once by semantic translation · Boolean, phrase and proximity syntax · citation extraction and chemical structure search · machine translation of full text via TAPTA · RSS alerts for new publications.

Strengths unmatched PCT and developing-country coverage; powerful translation. Limits variable OCR on older documents; some national collections lag.

USPTO

PatFTFull-text US granted patents, 1976 onward
AppFTPublished US applications
Patent CenterThe unified portal replacing legacy PAIR and EFOAS

Search modes: quick two-field query · advanced Boolean across title, abstract, claims, inventor, assignee and classification · CPC classification browse.

Strengths authoritative for US filings; rich legal-status data; Patent Center gives full prosecution histories — office actions and responses. Limits US only; pre-1976 patents are image-only.

The prosecution history in Patent Center is the underused resource: it shows you exactly which arguments persuaded an examiner to allow a claim in your field.
CSEG3060 · Unit VI · Lecture 3 — Patent Information, Databases and Administration9

Databases · 2 of 3

EPO Espacenet and the Indian patent database

EPO Espacenet

Free access to over 140 million patent documents from more than 100 countries.

Smart search for natural-language queries · advanced search across IPC, CPC, applicant, inventor, date range and priority number · interactive CPC/IPC browser · citation graph and INPADOC legal events, visualising family relationships and prosecution history across jurisdictions · Patent Translate supporting 32+ languages · Open Patent Services (OPS), a REST API for bulk retrieval, heavily used in patent analytics.

Strengths best-in-class global coverage, family and legal-event data. Limits some bulk-data features need a free EPO account.

IP India — IPAIRS

Provided by the CGPDTM: the Indian Patent Information Retrieval System, with a public search portal.

Search by application number, patent number, title, abstract, claims, full text, inventor, applicant, and CPC.

Covers applications filed in India under the Patents Act, 1970, including PCT applications entering the Indian national phase. Status tracking across Filed, Published, Under Examination, Granted, Withdrawn, Abandoned, Revoked.

Strengths authoritative and current legal status for Indian filings. Limits less polished interface; variable OCR on older scans; English-only.

The OPS REST API is what makes automated landscape analysis possible. If your project involves patent analytics at all, that is the endpoint you will build against.
CSEG3060 · Unit VI · Lecture 3 — Patent Information, Databases and Administration10

Databases · 3 of 3

Four databases, and what each is best for

FeatureWIPO PATENTSCOPEUSPTO PatFT/AppFTEPO EspacenetIP India (IPAIRS)
Geographic scopeGlobal — PCT + nationalUSA onlyGlobalIndia only
Document count100+ million12+ million140+ million~700,000 published
ClassificationIPC, CPCCPC (formerly USPC)CPC, IPCIPC, CPC
Full-text searchYesYes, 1976+YesYes
Machine translationYes — TAPTALimitedYes — Patent TranslateNo
API accessLimitedYes — Patent CenterYes — OPSNo
Legal status dataPartialComprehensiveExcellent — INPADOCAuthoritative for India
Best forPCT filings, global coverageUS prosecution historyComprehensive prior-art searchesIndian filings
The Best for row is the practical answer: start at Espacenet for prior art, go to Patent Center for prosecution arguments, use PATENTSCOPE for PCT and non-English filings, and check IPAIRS for anything that must be enforceable in India.
CSEG3060 · Unit VI · Lecture 3 — Patent Information, Databases and Administration11

Searching · 1 of 3

Three steps, and two complementary strategies

1 · Topic identification

Define the technical subject with precision. Vague topics produce unusable result sets.

2 · Query formulation

Combine keywords, classification codes and applicant filters.

3 · Result refinement

Apply date ranges, jurisdictions and citation analysis.

StrategyStrengthsWeaknesses
Keyword searchDiscovers synonyms and emerging terminology; intuitiveMisses translations and unusual phrasings; subject to noise
Classification searchReliable across languages and synonyms; preciseRequires familiarity with IPC/CPC; may miss innovative cross-domain applications
The most effective searches combine both — classification codes to narrow the field, then keywords to refine within it. Keyword-only searching is why students conclude an idea is novel when it is not; classification-only searching is why they miss the cross-domain application that anticipates them.
CSEG3060 · Unit VI · Lecture 3 — Patent Information, Databases and Administration12

Searching · 2 of 3

Syntax, and a worked query

Extended Boolean logic

AND, OR, NOT — standard operators
Quotation marks — exact phrase
Parentheses — nested grouping
Wildcards — typically * for zero or more characters, ? for a single character

Espacenet field syntax:

(cla = ("machine learning" OR "neural network")) AND ap = ("Google")

Searches claims mentioning machine learning or neural networks, in patents assigned to Google.

Worked example — AI-based intrusion detection

1 · Identify the CPC codes

G06N 3/04, G06N 3/08 — neural networks
H04L 63/00, H04L 63/14 — network security
G06F 21/55 — malware detection

2 · Construct the query on Espacenet

CPC = (G06N3/04 OR G06N3/08) AND (H04L63/00 OR G06F21/55) Date range: last five years Jurisdiction: US, EP, WO

3 · Refine with keywords such as “intrusion detection” or “anomaly detection” in title and abstract.

4 · Analyse — export bibliographic data, generate citation graphs, identify top assignees (Cisco, IBM, Palo Alto Networks), observe temporal trends.

CSEG3060 · Unit VI · Lecture 3 — Patent Information, Databases and Administration13

Searching · 3 of 3

Citation and family analysis — reading the graph

Backward citations

References cited by the patent. Useful for finding earlier art — the examiner and the applicant have already done part of your search for you.

Forward citations

Patents that later cite this one. Useful for assessing influence and mapping downstream innovation — a heavily forward-cited patent is a foundational one.

Patent families

The same invention filed in multiple jurisdictions. Useful for understanding a competitor’s global protection strategy and for identifying every relevant legal event.

Worked landscape — transformer architectures in NLP

Since Vaswani et al., Attention Is All You Need (2017), transformers have reshaped natural language processing. A landscape study reveals the commercial dimension of that academic breakthrough.

Databases: Espacenet and PATENTSCOPE. Query:

CPC = (G06N3/045 OR G06N3/0464 OR G06N3/0455) AND title/abstract = ("attention mechanism" OR "transformer" OR "self-attention")
Family analysis answers a question no single patent can: where does this competitor think the money is? A family filed in the US, EP, CN, JP and KR signals a global product; one filed only in India signals something else entirely.
CSEG3060 · Unit VI · Lecture 3 — Patent Information, Databases and Administration14

Part 2 begins

Who actually runs the patent system?

The first half of this lecture was about finding patents. The second half is about the institutions that create them — how an application is prosecuted, what happens after grant, and how the system has been reformed over the past two decades.

Tier 1

National patent offices
Examine applications under domestic law

Tier 2

Regional systems
Consolidate filing for groups of countries

Tier 3

International coordination
Treaty secretariat and standard-setter

The distinction that matters: Tiers 1 and 2 grant patents. Tier 3 does not. WIPO administers treaties and sets standards, but it has never granted a patent in its life — which is the same point made about the PCT in Unit V.

CSEG3060 · Unit VI · Lecture 3 — Patent Information, Databases and Administration15

Administration · 1 of 3

Three tiers, and the offices that matter

Tier 1 — national offices

USPTO — the largest national office by volume
EPO — supranational, examining under the EPC for contracting states
JPO — Japan
CNIPA — China, the fastest-growing filer globally
KIPO — Korea
IPO India — operating through four regional offices at Delhi, Mumbai, Chennai and Kolkata under the Patents Act, 1970

Tier 2 — regional systems

Consolidate filing procedures for groups of countries:

EAPO — Eurasian Patent Organization
ARIPO — African Regional Intellectual Property Organization
OAPI — African Intellectual Property Organization
EPO — acting as a regional grant body

Tier 3 — international coordination

WIPO, headquartered in Geneva, administers the PCT, the Madrid System for marks and the Hague System for designs.

WIPO functions as a treaty secretariat and normative standard-setter — not a granting authority.

Note that the EPO appears in two tiers. It examines under the EPC as a supranational office, and acts as a regional grant body for its contracting states — which is why a “European patent” is still validated nationally.
CSEG3060 · Unit VI · Lecture 3 — Patent Information, Databases and Administration16

Administration · 2 of 3

Six steps to grant, and what follows

1 · Filing

Specification, claims, abstract and drawings, to the office of first filing.

2 · Formalities examination

Verification of administrative requirements.

3 · Search

Identification of prior art; produces the search report.

4 · Substantive examination

Assessment against the patentability criteria.

5 · Office actions and responses

Exchanges between examiner and applicant, usually involving claim amendments.

6 · Grant or refusal

The culminating decision. Refusals may be appealed.

Patentability as a logical conjunction

Patentability = N ∧ IS ∧ IA ∧ SD

N novelty · IS inventive step (non-obviousness) · IA industrial applicability (utility) · SD sufficient disclosure. Each condition must be independently satisfied — which is what a conjunction means, and why examiners attack whichever is weakest.

Post-grant administration

Renewal and annuity payments, escalating over the 20-year term · opposition under Art. 99 EPC or s.25 of the Indian Act · invalidation and revocation · licensing registers · patent information and databases.

CSEG3060 · Unit VI · Lecture 3 — Patent Information, Databases and Administration17

Administration · 3 of 3

Two decades of reform, on three themes

Acceleration and quality

USPTO Track One — prioritised examination under the Leahy-Smith America Invents Act, 2011, for an additional fee, targeting final disposition in 12 months.

Patent Prosecution Highway networks — bilateral agreements letting offices fast-track applications using a partner office’s work product, cutting duplication and search cost.

EPO “Raising the Bar” (post-2010) — tightened examination standards, expanded supplementary European search reports.

Software subject matter

The most contested category. Computer programs as such were historically excluded in India (s.3(k)) and Europe (Art. 52 EPC) — but technical applications of software are routinely patentable.

USPTO Revised Subject Matter Eligibility Guidance (2019, updated through 2024) — clarifies Alice/Mayo with worked examples in AI.

EPO Case Law Compendium — continuous refinement of “technical character”.

Indian CRI Guidelines (2018, 2024 drafts) — computer-related inventions patentable where they show a technical effect or advancement.

Emerging technology and digital transformation

AI-assisted prior-art search — CNIPA and JPO have deployed ML tools to help examiners find documents.

SEPs and FRAND — India’s DPIIT issued a draft Standard Essential Patents Policy in 2021; the European Commission has advanced FRAND frameworks.

AI as inventor — the DABUS cases (Thaler v. Comptroller, EPO Boards of Appeal 2021–22; Thaler v. Vidal, US Federal Circuit 2022) rejected AI systems as inventors.

Digital — WIPO ePCT and the Digital Access Service; USPTO Patent Center replacing PAIR; EPO opposition by videoconference (2021); Indian Patents (Amendment) Rules 2020 and 2024.

The DABUS line is the one to watch: every major jurisdiction has now held that an inventor must be a natural person — a holding under active academic and policy pressure as AI systems contribute more of the inventive step.
CSEG3060 · Unit VI · Lecture 3 — Patent Information, Databases and Administration18

Consolidation

What you must be able to do by Lecture 4

Seven takeaways

  • 70% of technical information appears only in patent documents
  • INID codes let you read any patent front page in any language
  • The IPC is five levels deep; G06 and H04 cover most CS work
  • CPC adds ~200,000 codes, updates quarterly, and is used by EPO and USPTO
  • Search by classification and keyword together — neither alone is defensible
  • Administration runs in three tiers; only tiers 1 and 2 grant patents
  • Patentability = N ∧ IS ∧ IA ∧ SD — a conjunction, so every element must hold

Exam focus

  1. Decode an IPC code level by level, as on slide 5
  2. Four differences between IPC and CPC
  3. The four databases compared — scope, classification, legal status, best use
  4. Keyword versus classification searching: strengths and weaknesses
  5. The six prosecution steps and the patentability conjunction
  6. Post-grant administration — five activities
  7. Recent reforms on each of the three themes, with a named example

Do this before Lecture 4

Find the two or three CPC codes covering your group project. Run one Espacenet search combining them with a keyword. Note the top three assignees and whether anyone has already patented what you are building.

Where to start

Espacenet for prior art · Patent Center for prosecution history · PATENTSCOPE for PCT and non-English filings · IPAIRS for Indian enforceability

Next — Lecture 4: IPR in Software, Institutions and the Courts: copyright against patent for software, licensing models and open source, institutional IP policy and technology transfer offices, and the landmark cases.

1 / 15

Slides — click to jump Esc to close

Speaker notes