Monsanto · 2014–2015
Listening to the language of agriculture
An early initiative in text analytics, built around the words customers actually used.
Illustrative word network
What repeats. What connects.
Node area represents word count. Switch the edges to compare words appearing in the same response with examples of related meanings.
Example dataset · 12 invented Portuguese NPS comments from corn, soybean and cotton producers. No real customer data.
Choose a crop, then select a word to see the comments behind it.
Example input · producer comments
Synthetic Portuguese examples, created for this illustration. Counts use exact words after lowercasing and removing punctuation. Solid edge weight is the number of shared responses. Dashed semantic links are manually curated examples, not an embedding model, an inferred finding or the original Monsanto analysis.
NLP / Customer insights
From feedback to marketing insight
Start with open-ended NPS responses
Look beyond the score to what producers and customers describe in their own words.
Count words and compare texts
Use simple word-frequency and similarity analysis to discover relationships worth investigating.
Interpret language in context
Read findings alongside regional expressions, slang, tone and the different crop and brand contexts.
Make the findings useful to marketing
My reports helped marketing understand what customers were reporting, including tone, slang and regional language across Monsanto’s corn, soybean and cotton businesses and their brands.
Initiative before a formal data-science role
I saw an opportunity to apply computational methods to qualitative customer feedback while still an intern. The contribution was connecting a practical text-analysis approach with a business team’s need to understand its customers.
This retrospective is based on my account of the work. The original code, customer responses and reports are not published here. Tone was part of the interpretation; no automated sentiment-classification result or measured commercial lift is claimed.