Add WorldCom to the Concept Stop Word list to clean up cluster titles.

Title: When WorldCom Keeps Showing Up in Cluster Titles—and How to Fix It

If you’ve ever built clusters in Relativity and watched a single term steal the spotlight, you know the feeling. The titles that are meant to summarize ideas end up looking repetitive, with WorldCom popping up again and again. It’s not just annoying; it makes your clusters feel less useful, like a map where every street is named “Road.” So what can you do to rebalance the titles and let the real themes shine through?

Here’s the concise answer you’ll want to remember: add the client name, WorldCom, to the Concept Stop Word list. That single action tunes the clustering instrument so that common, non-substantive terms don’t drown out the meaningful concepts you’re trying to surface.

Let’s unpack why this is the right move, why other options don’t quite hit the mark, and how you can apply it smoothly in your environment.

The problem in plain terms

Cluster titles are meant to reflect the topics inside a group of documents. They should be concise, informative, and distinct from one another. When a term like WorldCom appears frequently, it tends to dominate the titles, even when it doesn’t help distinguish one cluster from another. The result is a set of titles that feel redundant and less insightful. It’s a classic case of noise obscuring signal.

Think of it like a speaker trying to tell you about a concert’s set list, but the mic keeps feeding you the loudest word in the song’s title. The rest of the band—concepts, topics, and subthemes—gets muffled. The fix isn’t more noise; it’s filtering out the noisy word so the underlying ideas can come through clearly.

What fixes it—and why this is the right move

A. Add the client name to the Concept Stop Word list.

This is the precise, targeted adjustment that reduces over-emphasis on a term that’s useful for context but not for differentiating clusters. Stop words are terms the clustering process should ignore when generating titles. By listing WorldCom as a stop word, you’re telling the system, in effect, “We hear you, WorldCom, but we don’t want you to count toward the title’s meaning.” The clusters then reflect the real topics—policies, contracts, project scopes, vendor relations, or risk areas—rather than the fact that one client’s name appears often.

Why the other choices don’t address the root cause

• B. Run a keyword expansion on the client name.

This sounds like it could widen the vocabulary, but it actually risks layering in more signals that aren’t helping with distinction. If your goal is clearer titles, expanding a term tends to pull in more instances of the same noise, not reduce it. It’s a bit like adding more synonyms to a paragraph that’s already crowded with filler words.

• C. Enable the email header filter on the conceptual index.

That might seem relevant, but it doesn’t directly tackle how cluster titles are formed. Filters on email headers are great for narrowing data sources or improving search relevance, but they don’t reweight or suppress terms in the conceptual model used for clustering.

• D. Re-run the cluster set including all documents.

So you want to refresh the clusters by re-processing everything? That can produce a different set of titles, but it won’t guarantee the WorldCom term will fade from the titles. If the root issue is how the model treats that term, you’ll just get different—but still noise-heavy—titles. Without removing the term from consideration, you’re chasing the tail.

A practical how-to: applying the Concept Stop Word fix

If you’re ready to implement, here’s a straightforward path you can follow. It’s light on steps but heavy on impact, and you can adapt it to your workspace’s specific UI:

1. Locate the Concept Stop Word list.

• In Relativity, settings for concept labeling and clustering often live in the analytics or indexing modules. Find the area that governs stop words or concept filters. If you’re on a newer version, there may be a dedicated “Concept” or “Analytics” tab with a Stop Word List you can edit.

2. Add WorldCom (and its common variants).

• Enter WorldCom as a stop word. If your environment uses case-insensitive matching (which is common), you don’t have to worry about capitalization. Still, it’s wise to add likely variants, such as “World Com” or any other brand representations that appear in titles. The key is to capture all the forms that show up in cluster titles.

3. Save and validate on a sample.

• Don’t flip the switch on the entire workspace without a test. Apply the change to a representative subset of documents or a pilot project. Generate new cluster titles and compare them to the previous output. Are the titles now richer in themes like contract terms, vendor performance, or data governance? Do you still see WorldCom in the titles? If you still see it, you may need to broaden the stop word list slightly, or check for derivatives that slip through.

4. Review results for balance.

• A good sign is that cluster titles are varied and topic-focused rather than dominated by a single entity. If you notice other non-substantive terms stealing the spotlight, consider adding them as stop words as well. The goal is to accentuate the meaningful concepts without over-filtering.

5. Implement governance for stop words.

• Create a lightweight policy: which terms should be stop words, how often should you review the list, and who approves changes. This helps keep cluster quality high as new clients, vendors, or project names appear in your corpus.

A few practical tips to keep things smooth

• Start with the big offenders. If WorldCom is top of mind, you’ll likely discover other frequent entity names that show up in titles. Tackle them in small batches so you can measure impact without overfitting the model.

• Don’t overdo it. If you turn too many terms into stop words, you risk erasing useful distinctions. It’s a balance — stop words should tame noise, not erase signals.

• Test with real-world variability. Your dataset will evolve as new projects roll in. Schedule periodic reviews of the stop word list to keep it aligned with current work.

• Consider related governance angles. Metadata quality, naming conventions, and data governance practices all influence how clean clustering output remains. A small adjustment in one area can ripple positively through others.

A quick aside about the bigger picture

There’s a simple, human truth behind these settings: we want systems to reflect what matters, not what’s loud. In many information-heavy environments, client names, vendor labels, or project identifiers are necessary for context—but they aren’t the essence of the document content. The cluster titles should foreground themes like risk, scope, deliverables, or policy language. The Concept Stop Word list is one of those quiet, backstage tools that helps keep the spotlight where it belongs.

A few common-sense caveats

• If a term becomes essential in a specific domain, you may later remove it from the stop word list for particular projects. Flexibility matters.

• Always pair changes in the stop word list with a quick sanity check: do the clusters still feel coherent? If you notice new oddities, adjust thoughtfully.

• Documentation helps a lot. When you record why you added a word to the stop list, others can understand and replicate your decisions.

What this means for your work

The way you manage cluster titles can make a real difference in how easily teams find, interpret, and act on the results. A clean set of cluster titles acts like a well-organized filing cabinet: you know where to look, and you can trust the labels to guide you toward the right content. By selectively removing terms that don’t contribute to topic differentiation, you free up space for the real ideas to surface.

If you’re currently wrestling with noisy cluster titles, the simplest rescue is often the most effective: include WorldCom in the Concept Stop Word list. It’s a small adjustment with a meaningful payoff—one that can sharpen the clarity of your analysis and save you time down the line.

A few closing reflections

• You’ll likely encounter this issue in other contexts, too. Stop words aren’t only about Clustering; they can improve search results, document tagging, and even how you present summaries to stakeholders.

• Treat stop word management as an ongoing practice, not a one-off fix. Data, terms, and project names shift over time, and your toolkit should shift with them.

• The best solutions come from a balance: keep the focus on substantive concepts, but don’t be afraid to refine your approach as you learn what your audience needs from the titles.

TL;DR

When a client name dominates cluster titles, add that name to the Concept Stop Word list. It cools the noise, letting meaningful themes take center stage. B and D may miss the mark. C doesn’t directly tackle title-generation dynamics. A targeted stop word adjustment, tested on a sample, can dramatically improve the clarity and usefulness of your clustering output.

If you want to keep the conversation going, I’m curious: what other terms tend to pop up too often in your cluster titles, and how have you addressed them so the themes stay crisp and helpful?