clean_up_text
Clean Up Text
Removes JavaScript, CSS tags, JSON, and other markup from text, returning pure decoded plaintext. Useful for preprocessing HTML or markup-heavy content before analysis.
clean_up_text
Removes JavaScript, CSS tags, JSON, and other markup from text, returning pure decoded plaintext. Useful for preprocessing HTML or markup-heavy content before analysis.
translate_text
Translates text between supported languages. Supports auto-detection of the source language. If source and target languages are the same, paraphrasing is applied instead. Handles obfuscated and slang content, preserving emotional register — profanities and slurs are translated faithfully.
detect_language
Detects the languages used in a text fragment and returns a breakdown by offsets. Useful for identifying the language(s) of unknown text or multilingual content where different parts may be in different languages.
list_languages
Retrieves all languages supported by Tisane, with metadata including native name, English name, script direction, and encoding. Use this to discover valid language codes for other tools.
lookup_word
Looks up a word in Tisane's language models to retrieve its senses (meanings), definitions, and crosslingual family IDs. Can also retrieve inflected forms, hypernyms (broader terms), and hyponyms (narrower terms) for a given family. Useful for linguistic research, understanding word senses, and exploring semantic relationships.
semantic_similarity
Calculates the semantic similarity between two text fragments, returning a score between 0 (completely different) and 1 (identical meaning). Supports cross-language comparison, allowing you to compare texts written in different languages.
compare_entities
Compares two compound named entities and identifies whether they refer to the same entity, with specific differences if they don't match. Supports cross-language comparison. Currently supports person entities only. Useful for name matching, deduplication, and identity resolution across different languages and naming conventions.
analyze_text
Performs comprehensive NLP analysis on text including abusive content detection (hate speech, cyberbullying, sexual harassment, criminal activity), sentiment analysis, named entity extraction, and topic detection. Supports 30+ languages. Configure the analysis through optional settings to enable/disable specific features like document-level sentiment, word-level tokenization, parse trees, and explanations.
Analyze text to detect abusive content (hate speech, cyberbullying, sexual harassment, criminal activity), extract sentiment, named entities, and topics across 30+ languages. Detect languages in text fragments, calculate semantic similarity between texts, compare named entities cross-linguistically, translate text between supported languages, and clean up markup from text. Provides low-level NLP capabilities including tokenization, part-of-speech tagging, and morphological analysis. Supports configurable content moderation with severity scoring and explanations, document-level and expression-level sentiment analysis, and direct access to underlying language model entries.
Common questions about connecting Tisane to AI agents with Metorial.