Text & Sentiment Analytics

Text sources gathered and cleaned, a theme codebook and sentiment labels you can defend, checked models, weekly readouts and the routines that turn reviews, tickets and comments into issues someone fixes.

Free · Mac (Apple silicon) and Windows · v1.0.5 · Other platforms

Then open the Marketplace in the app, search for Text & Sentiment Analytics, and click Install. It goes into your own files, and works offline from then on.

What it installs.

1

Pillar

1

Area

50

Projects

190

Starter tasks

📊 Data & Analytics

Text & Sentiment Analytics

Analysing reviews, tickets, survey comments and other text at scale to find themes, sentiment and emerging issues.

  1. Inventory of text sources and their owners
  2. Decisions the text analysis must inform
  3. One clean table of all feedback text
  4. Redacting personal details from feedback text
  5. Cleaning and deduplication rules for feedback text
  6. Reading 200 comments by hand before any model
  7. First theme codebook from a hand-coded sample
  8. Sentiment label definitions and edge-case rules
  9. Baseline theme and sentiment snapshot for the last 90 days
  10. Weekly theme and sentiment readout
  11. Monthly codebook upkeep
  12. Double coding and agreement checks
  13. Model accuracy spot-checks against human labels
  14. Emerging issue watch for new terms and spikes
  15. Routing themes to owners and closing the loop
  16. Verbatim quote bank
  17. Quarterly theme trend report for leadership
  18. Volume and coverage checks on text feeds
  19. Annual review of sources, codebook and models
  20. Tokenisation, stop words and n-gram basics
  21. Lexicon sentiment and where it breaks
  22. Topic modelling with LDA and embedding clusters
  23. Supervised theme classification from labelled examples
  24. Aspect-based sentiment analysis
  25. Using a language model as a coder, with checks
  26. Thematic analysis methods from qualitative research
  27. Text analytics reading habit
  28. Choosing a text analytics platform or building in-house
  29. Defining the sentiment metric used in reports
  30. Who writes comments and who stays silent
  31. Linking themes to refunds, churn and repeat contact
  32. Handling feedback in several languages
  33. Fixing the support ticket tagging scheme
  34. Shrinking the uncategorised share below ten percent
  35. Feedback sprint for a product launch
  36. Comment analysis after an outage or incident
  37. Annual voice of the customer report
  38. Readout workshop with product and support teams
  39. Tracking reaction to a price or policy change
  40. Ethics and terms check for public social media text
  41. Theme analysis in a spreadsheet without code
  42. Ten thousand comments due by Friday
  43. Combining interview transcripts with large-scale comment themes
  44. Analysing public consultation responses
  45. Fine-tuning a domain sentiment model
  46. Semantic search across all feedback
  47. Sentiment accuracy across dialects and code-switching
  48. Error analysis of sarcasm, negation and mixed sentiment
  49. Change-point detection on theme volumes
  50. Model card for the text analytics pipeline

Where it starts.

Each project arrives with its purpose, milestones and first tasks written out. Here is how the first three read.

Inventory of text sources and their owners

Customer words are usually scattered across an app store console, a review site, the helpdesk, survey exports, live chat and a social listening tool, each owned by a different team. Listing every source with its owner, volume per month, history available and export route shows what can be analysed together and what is missing. Without it, the first theme report quietly covers half the feedback and nobody notices.

First tasks

  • List every place customers or staff write free text about the product or service
  • Ask each source owner for monthly volume and how far back the history goes
  • Record the export route and any rating field for each source
  • Pick the three sources to analyse first and note why

Decisions the text analysis must inform

Theme reports that answer no particular question get read once and filed. Writing down the four or five decisions the analysis should feed, such as which bugs to fix first, whether a price change landed badly or which help articles to rewrite, sets the level of detail, the refresh rhythm and who must see each finding. It also gives you a polite reason to say no to vanity word clouds.

First tasks

  • Ask product, support and marketing leads which decision they wish customer comments could settle
  • Write each decision with its owner and the date it is next made
  • Note the sources and cuts each decision needs
  • Send the list to stakeholders and record their agreement

One clean table of all feedback text

Analysing reviews in one file, tickets in another and survey comments in a third makes cross-source themes impossible and doubles every piece of cleaning. A single table with the same columns for every source, such as record id, source, timestamp, text, rating, product, language and customer key, is the base every later model and report reads from.

First tasks

  • Draft the shared column list for the combined feedback table
  • Map each source's fields onto the shared columns
  • Load the first three sources and compare monthly row counts with the source systems
  • Schedule the refresh and show the last load time beside the table

Make it yours.

Install Text & Sentiment Analytics, then tell Aurum the date you're aiming for. It checks whether the plan fits the weeks you actually have.

Free · Mac (Apple silicon) and Windows · v1.0.5 · Other platforms

Blueprints are general information to help you organise your own work. They are not medical, financial, legal or other professional advice. See the Terms.