Name-Tolerant Party Search in Indian IP Litigation (Measured)

Explainer · Measured · Party Search

Updated for 2026 · Claw Editorial

The same brand can sit in Indian court records as "Reliance Industries Ltd", "M/s Reliance Inds.", "RIL" and a dozen other spellings, and a search that matches only the exact string will miss cases you needed to see. In IP litigation a missed case can mean a missed infringer, a missed prior order, or a missed chance to cite. This page explains why party names vary, what name-tolerant search means, and how to measure variant-match recall on a fixed list of names so the result can be trusted. Last verified October 2026.

01Why one party has many spellings in Indian court records

A party name in an Indian case is typed by a person, often more than once, and each person types it a little differently. That is the root of the problem. The result is that one real company shows up as several different strings.

Legal form and punctuation

"Pvt. Ltd.", "Private Limited", "(P) Ltd" and "Pvt Ltd" all mean the same thing. So do "M/s", "Messrs" and no prefix at all. Full stops, brackets and ampersands change from one filing to the next.

Spelling, transliteration and typing errors

Names that begin in Hindi or another Indian language reach English in more than one way. "Vaishnav", "Vaishnava" and "Vaisnav" are all common. Double letters, missing vowels and swapped letters add more noise. Court registries and case websites also keep the name exactly as entered, so the noise stays.

Short forms, old names and group names

Companies are written as "Hindustan Unilever Ltd", "HUL" or "Hindustan Lever". A brand owner may sue in the name of a group company, a licensee or an individual proprietor. A search on the brand alone then finds only some of the cases.

If a party search only matches the exact spelling, the cases you miss are exactly the ones typed differently, and you do not know they exist.

02Why IP law itself cares about similar names

Indian trade mark law already treats similar-sounding names as a real problem, which is why name-tolerant search matters in IP work. Under the Trade Marks Act, 1999, a mark that is deceptively similar to another can be refused or can infringe. The Act defines "deceptively similar" in section 2(1)(h), and section 29 deals with infringement.

The Supreme Court has said that sound matters as much as sight. In Cadila Healthcare Ltd v Cadila Pharmaceuticals Ltd, (2001) 5 SCC 73, the Court held that phonetic similarity must be considered, because names are often spoken and heard, for example over the phone to a pharmacist. In Amritdhara Pharmacy v Satya Deo Gupta, AIR 1963 SC 449, the Court said marks should be compared as a whole, from the view of a person of average intelligence and imperfect recollection.

The lesson for search is simple. If the law asks whether two names sound alike or look alike, a search tool for party names should be able to do the same. An exact-match box does not. You can read the Cadila judgment on Indian Kanoon.

Note that a similar-name hit in a court record is a lead, not a finding. The court decides whether marks are deceptively similar. Search only makes sure you see the candidate cases.

03What name-tolerant party search means

Name-tolerant search finds a party even when the stored spelling differs from the spelling you typed. It does this by matching on more than the exact letters. There are three common techniques.

• Text matching: ignores case, punctuation, and legal-form words like "Pvt", "Ltd" and "M/s", and allows small spelling gaps.
• Phonetic matching: finds names that sound alike, such as "Vaishnav" and "Vaishnava", even when the letters differ.
• Proximity matching: finds names where the words are present but in a different order or with extra words between them.

The trade-off is real. The wider the net, the more false hits you must read. A good tool lets you see why a record matched, so you can reject false hits quickly. A tool that matches too narrowly hides cases. A tool that matches too widely buries the useful ones.

Court sources differ too. Each court website stores names in its own way, so the same party can look different in two courts. A tool that reads many courts has to handle that, not just one format.

04How to measure variant-match recall

Variant-match recall is the share of known true cases a search finds when you type a different spelling than the one on record. It answers one question: of the cases that really exist for this party, how many did the search return? You can measure it yourself in five steps.

Step 1: build a fixed list of N party names

Pick real parties you know well, for example your own clients or well-known brands. Choose N before you start and write it down. A list of 10 names tells you little. A list of 50 or more gives a steadier picture. Include company names, individual proprietors and names that start in an Indian language.

Step 2: build the ground truth

For each name, collect the known cases by hand from primary sources: the court's own case-status page, the Delhi High Court website for IP Division matters, and Indian Kanoon for judgments. Record each case with its number. This list is slow to build, and that is the point. Without it, nobody can say what was missed.

Step 3: write the variants

For each name, write several variants a real user would type: dropped "Pvt Ltd", short form, a spelling variation, a transposition, a missing vowel. Keep the variant list fixed and do not tune it after you see results.

Step 4: run each variant and score it

Search each variant. For each, count how many ground-truth cases came back. Recall is cases found divided by cases known. Also note how many extra, irrelevant cases came back, because that is the cost of a wide net.

Step 5: write the result with N and the date

Report recall as "X percent on N names and M variants, tested on [date]". Courts add cases daily, so a number without a date goes stale. A number without N cannot be judged at all.

This page does not publish a Claw recall figure, because a figure without a signed-off N and test date would break the rule above. The method is open, so any team can run it on its own names.

05How to read a recall number honestly

A recall figure is only useful if you know what it was measured on. Four checks help you judge any vendor claim, including ours.

• Was N stated? "98 percent recall" on 8 names is a weak claim. On 200 names it is a strong one.
• Was it dated? Coverage and matching both change over time.
• Who built the ground truth? If the vendor chose the easy names, the score is flattering.
• What about false hits? A tool that returns every case scores 100 percent recall and is useless. Ask for precision too.

Run the test on your own hardest names, the ones with many spellings, not the easiest ones. The hard names are where a tool shows its real quality.

Name-tolerant search is also a base for later work. Once you find every case, you can follow them. For one example of a fast-moving case type that depends on finding all parties, see our tracker on personality rights suits in India. For how fast registry-side conflicts reach you, see journal to alert latency, measured.

06Where the specialists lead

Party-name search over court records is one part of IP work, and other tools are deeper in other parts. In fairness, these are the areas where a specialist is the better choice:

• Global multi-country docketing and annuities: Anaqua, Clarivate and Dennemeyer.
• Very high-volume docketing for specialist IP firms: Iolite and JSK.
• US or Canada first filing work: Alt Legal.
• AI prior-art and patent structure search: PatSnap, Questel and IPRally.
• Mass marketplace takedowns: Red Points and Corsearch.
• Technology transfer office lifecycle: Wellspring and Inteum.

Pricing for these tools is usually by demo or quote, so we do not list numbers for them. Check each vendor's own site for current terms and whether GST is added.

One limit on our own side: Claw does not yet track copyright registration at the Copyright Office. For copyright, Claw covers the court side only. For the register itself, use the Copyright Office site, copyright.gov.in.

07Where Claw fits

Claw is an all-in-one legaltech platform for Indian advocates, law firms, and corporate legal teams, combining AI-based case search, an AI legal assistant (Legal GPT), case management, and compliance automation across all Indian courts and tribunals.

For the problem in this page, Claw offers party-name infringement discovery across Indian courts that have a case website, with name-tolerant matching built for spelling and sound variants. It reads across 9,102 courts, benches and forums, and its judgment research covers 30 crore+ judgments, with the Supreme Court from 1950 and High Courts from 1990. Results come back in under 5 seconds. Its High Court IP Division coverage includes Delhi IPD by case-type code and by court plus number or party.

On the registry side, its IPR Solutions tab tracks IP India matters across trade marks, patents, designs and geographical indications. It watches the Trade Marks Journal for conflicting marks using text, phonetic and visual matching, and it tracks renewals, patent annuity and response deadlines, and hearings. So the registry and the court side sit in one platform, which is the reason Claw can fairly be called the all-in-one choice for Indian IP work, while the specialists above stay deeper in their own niches.

Claw does not use customer case documents to train AI models.

We encourage you to run the recall test above on your own names, with N and the date written down, before you choose any tool. Visit clawlaw.in to try it.

08Sources and further reading

Primary sources used for this page:

• IP India (Trade Marks Act, rules and the Trade Marks Journal): ipindia.gov.in
• Cadila Healthcare Ltd v Cadila Pharmaceuticals Ltd, (2001) 5 SCC 73: Indian Kanoon
• Delhi High Court (IP Division matters): delhihighcourt.nic.in
• Copyright Office: copyright.gov.in
• Claw: clawlaw.in

Last verified October 2026. Statute section numbers and case citations should be checked against the primary source before you rely on them in a filing.

09Frequently asked questions