Company Name Cleaning Combines Brand Normalization Rules, Phonetic Matching and AI

·1 views

A developer describes a three-part method for cleaning messy company names in databases: normalization rules, phonetic algorithms, and a language model for ambiguous cases. The approach was first built around 2.3 million names gathered from ad exchanges, CRM exports and forms, where no validation existed at entry. Because the same firm can be written many ways, exact matching fails, so comparing how names sound works better than comparing spellings. The method relies on lists of legal-form suffixes and filler terms to strip as noise, applied in a strict order: standardize casing and punctuation, replace longest terms first, and use word boundaries.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Octomind: open-source Rust AI agent harness aims to reduce token usage

Octomind is an open-source AI harness written in Rust, shared on Hacker News with a link to its GitHub repository. The project is positioned as a way to save tokens when working with AI models. The Hacker News post collected 11 points and 2 comments at the time it was listed. The source material provides only the repository link and engagement figures, so details on specific features or benchmarks are not confirmed here.

0 comments Read more at Hacker News

ProgrammingDEV Community ·

OnePassApps' Ask Sam panel opens apps by command and works without an account

A developer working on OnePassApps examined the "Ask Sam" button found on every page of the app suite. The panel appears beside the page and offers a text composer, with chips that pre-fill a command rather than submitting it. Typing "Open" followed by an app name, such as "Open Image Studio", navigates straight to that app, and this works while signed out. Free-form requests receive a fixed guidance message explaining the command syntax instead of an AI-generated answer. Network logging during an app command showed only the app's own configuration fetch, with no chat request sent.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

SoundHound to Report Q3 Results November 5, With Options Pricing About a 12.5% Move

SoundHound AI (SOUN) is scheduled to report its third-quarter results on November 5, with its investor call at 5 p.m. Eastern. Analyst consensus as of October 29 called for revenue of about $62.2 million and a per-share loss of $0.05. The company's 2026 revenue guidance of $230 million to $260 million was set in August, but its September 4 acquisition of LivePerson, which booked $52.5 million of revenue in the second quarter, is not reflected in that consensus. Options pricing implies a move of roughly 12.5% around the report, and short interest stood at 170.8 million shares as of October 15.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Fine-Grained Password Policies: How to Set Them Up in Active Directory

This guide explains how to create Fine-Grained Password Policies (FGPP) in Active Directory, which allow different password and account lockout rules for specific users or groups within one domain. FGPPs can only be applied to user objects and global security groups, and the domain functional level must be Windows Server 2012 or later. Administrators need Domain Admins membership and either the Active Directory Administrative Center or the Active Directory PowerShell module. The article walks through creating a policy in the Password Settings Container, noting that lower precedence values take higher priority. Finally, it shows how to verify which policy applies to a user using the Get-ADUserResultantPasswordPolicy PowerShell command.

0 comments Read more at DEV Community