Model / text · All operations
Extract meaning from text
Sentiment, named entities, topics, language detection.
Status
- In arcform
Not built yet
arcform does not do this yet. Its place on the roadmap
- Shape of the result
- one row out per row in, with columns added or changed
- Level
- needs an arcform operator
Variants
- Pick out named entities
Find the people, places, organisations and other named things mentioned in each text.
- Find the topics in a body of text
Discover topics across a collection of texts and say which topics each text belongs to.
Extract fields from documents
Read invoice line items or key-value pairs out of scanned or structured documents.
Score or label each text
Add a sentiment score, language, class label or summary to each text value.
Other text and media analysis
Analyses of text or media that fit none of the other variants, such as acoustic features of audio.
What other tools call it
Meridian's classification; an independent blind rating agreed on 86.2% of 312 rows (exact class, version 2).
Alteryx Designer (7)
Cognitive Services Text Analyticsread 2026-09-27Call Azure Cognitive Services for sentiment, key phrases and language detection
Invoice Extractorread 2026-09-27Extract line items and metadata from invoice files, backed by an LLM connection
Key-Value Pair Extractionread 2026-09-27Find key-value pairs in documents with Tesseract OCR and fuzzy matching
Named Entity Recognitionread 2026-09-27Identify people, places and other entities in text with spaCy
Sentiment Analysisread 2026-09-27Score text as positive, negative or neutral
Text Summaryread 2026-09-27Summarise bodies of text
Topic Modelingread 2026-09-27Identify and categorise topics in a body of text
Apache Hop (1)
Detect languageread 2026-09-27Identify the natural language of a text field
KNIME Analytics Platform (2)
BERTopic Extension (extension)read 2026-09-27Extension outside the base install: 1 current node: Topic Extractor (BERTopic)
KNIME Audio Processing (extension)read 2026-09-27Extension outside the base install: 8 current nodes: Acoustic Feature Extractor (legacy); Audio Data Extractor (legacy), …
OpenRefine (1)
detectLanguage(s)read 2026-09-27Return the locale code of the language a string is written in
RapidMiner Studio (7)
Dictionary-Based Sentiment (Documents)read 2026-09-27Build a sentiment model from a list of words with positive or negative weights
Extract Sentimentread 2026-09-27Score the sentiment of a text column using sentiment dictionaries or an external API
Extract Topics from Data (LDA)read 2026-09-27Find topics in a text column by latent Dirichlet allocation
Extract Topics from Documents (LDA)read 2026-09-27Find topics in a collection of documents by latent Dirichlet allocation
Text Classificationread 2026-09-27Apply a Huggingface text classification model (such as sentiment) to a text column
Token Classificationread 2026-09-27Apply a token classification model (such as named entities) to a text column
Zero Shot Classificationread 2026-09-27Classify text into user-given candidate labels with a zero-shot model
Background
Information extraction (IE) is the task of automatically extracting structured information from unstructured and/or semi-structured machine-readable documents.[1] A broad goal of IE is to allow computation to be done on previously unstructured data.[1] Sentiment analysis, also known as opinion mining, is used to systematically identify, extract, quantify and study affective states and subjective information.[2]
When to use it
Sentiment analysis is widely applied to voice of the customer materials such as reviews and survey responses, to online and social media, and to healthcare materials.[2] The accuracy of a sentiment analysis system is, in principle, how well it agrees with human judgments.[2] Research has found that human raters typically agree only about 80% of the time when judging the sentiment of a text.[2] A computer system doing sentiment analysis will have trouble with negations, exaggerations, jokes or sarcasm, which are typically easy for a human reader to handle.[2] Read a sample of texts beside their scores before a sentiment column goes into a report, and keep the sample with the result. (Meridian's view)
Alteryx recommends not running its Text Pre-processing tool before its Sentiment Analysis tool, because pre-processing can remove features the sentiment tool relies on.[3] When Alteryx's Sentiment Analysis tool finds sentiment at sentence level, it uses punctuation to split the text into sentences, so the text must contain end punctuation.[3] Score sentiment on the text as it was written, before any cleaning step strips its punctuation or words. (Meridian's view)
Alteryx's Sentiment Analysis tool can only analyse data containing English text.[3] OpenRefine's detectLanguage function returns a string locale describing the language of a string, so "Hello, world!" gives "en".[4] Where a column mixes languages, detect the language of each value first and send each language to a model built for it. (Meridian's view)
The method
Information extraction typically involves processing human language texts by means of natural language processing (NLP).[1] Information extraction dates back to the late 1970s, in the early days of NLP.[1]
A basic task in sentiment analysis is classifying the polarity of a text at the document, sentence or feature/aspect level, as positive, negative or neutral.[2] Beyond polarity, sentiment classification can look at emotional states such as enjoyment, anger, disgust, sadness, fear and surprise.[2] Precursors to sentiment analysis include the General Inquirer, which provided hints toward quantifying patterns in text.[2]
RapidMiner's Text Classification operator applies a model trained to predict the most likely class of a text, such as whether its sentiment is positive or negative.[5]
Sources
- [1]Information extraction - Wikipediasecondary source
- [2]Sentiment analysis - Wikipediasecondary source
- [3]Sentiment Analysisdocumentation
- [4]GREL functions | OpenRefinedocumentation
- [5]Text Classification - Rapidminer Documentationdocumentation