This word cloud was created in Voyant Tools from a dataset of recovered items. To prepare the data, the item-name column was separated from the rest of the spreadsheet and converted into a clean text list for analysis. The aim was to ensure the visualization reflected item types and meaningful descriptors rather than packaging language, transcription artifacts, or product-label conventions.

​

Because word clouds are based on raw word frequency, the text was cleaned in two passes to reduce high-frequency “noise.” The first pass removed non-descriptive terms common in product names (such as measurement words and generic placeholders like oz, ounce, gram, and brand). Once those were removed, a second pass eliminated frequent but low-information modifiers—especially size, color, and mixture/state terms (such as size, large, white, black, mix, and dry) that can dominate counts without adding much interpretive value.

Some multiword names were standardized with hyphens so they would appear as single terms in the cloud (for example, Nature Valley → Nature-Valley; Trader Joe’s → Trader-Joe’s), and transcription artifacts from voice dictation (such as uh) were also removed.

​

A small number of common words were intentionally kept when they served as shared head terms across many compound names. For example, butter appears often largely because of phrases like almond butter, peanut butter, and shea butter. Keeping butter visible preserves a real pattern in the dataset. After these steps, the cleaned text was entered into Voyant Tools to generate the final word cloud, highlighting recurring object types while reducing misleading frequency effects.