Text mining converts large volumes of unstructured writing into organized data for analysis. It combines natural language processing to structure language, machine learning to detect patterns or predict outcomes, and artificial intelligence to automate the process across varied sources. section 5.1
Use text mining to:
- Track sentiment and customer preferences. section 5.1
- Spot trends and identify potential threats. section 5.1
- Extract phrases, entities, relationships, sentences, and parts of speech for further analysis. section 5.1
Traditional techniques often approximate meaning through frequent phrases, keywords, or predefined dictionaries. Large language models instead process text sequentially and can connect information across passages, although their internal shortcuts and limitations remain difficult to assess. section 5.1.1
Customer feedback
A review-analysis workflow can examine post-purchase surveys to find positive aspects and improvement areas. The data may be disorganized, emotional, misspelled, and written in Portuguese; the dataset includes ratings, review titles and messages, identifiers, and timestamps. section 5.2.2
