Every case file has the same problem. The same person appears under different names. Initials in one document, a full name in another, a nickname in a witness statement, a different spelling in a company record. Two people with the same common name appear in the same case and must not be confused.
Get this wrong in one direction and the case misses a connection. Get it wrong in the other and two unrelated people are treated as one, which is worse.
Spelling is not enough
The easy approach is to match names by how similar they look. It fails both ways. "Ramesh Kumar Sharma" and "R. K. Sharma" look different and may be the same man. Two people both called "Mohammed Iqbal" look identical and may have nothing to do with each other.
Anvexer does not decide on spelling alone. It scores each possible match on several kinds of evidence together:
- how similar the names are
- whether both records are the same kind of thing, such as two people or two companies
- whether they appear in the same documents
- whether their details agree, such as date of birth, address or identity numbers
Two records with identical names but contradicting identity details are kept apart, or sent for review. They are not joined because the names match.
Three outcomes, not two
Every possible match falls into one of three groups.
Strong matches are shown as one person throughout the case. The investigator sees one profile, with all the name variants recorded on it.
Uncertain matches are listed for review. The investigator sees the two records side by side with the evidence for and against, and decides.
Weak matches are left as separate people.
The thresholds between these groups can be adjusted for a case.
The machine never merges permanently on its own
Automatic matching links records together; it does not destroy either of them. A permanent merge is always an investigator's decision, made on the review screen: the same person, merge them for good, or keep them separate.
When an investigator says two records are different people, that decision stands and the system does not override it. When an investigator creates or names a record themselves, their version is the one shown for that person.
Why this matters for the case
Identity is the foundation of everything else in a case file. The association graph, the timeline, the financial links and the dossier all depend on knowing who is who. A tool that merges too eagerly produces a confident, wrong network. A tool that never merges produces a fragmented one that hides the real connections.
Anvexer's approach is to do the clear cases automatically, show its reasoning on the unclear ones, and leave the final call with the person accountable for the case.
Frequently asked questions
Does Anvexer handle names written in different scripts?
Matching today works on the text as it appears in the documents. Investigators should review matches between names written in different scripts on the review screen.
Can a merge be undone?
Strong matches that are shown as one person are not permanently merged and can be separated by marking the pair as different people. A permanent merge made on the review screen collapses the records into one.
Does it match organisations as well as people?
Yes. The same approach applies to companies, organisations and other records in the case.
Analysis published by the Anvexer Research Cell at Sobitai Solutions LLP. Exploring AI decision-support, NATO/Admiralty evidentiary frameworks, and graph intelligence for modern law enforcement.