ceven

Tisane

Scans incoming customer messages for problematic content and sentiment, extracts core topics, and translates global feedback into a single language for your support team.

Try Tisane in Ceven

Ask Ceven anything
Standard

Why use Ceven?

  1. AI native Tisane integration

    • Describe the outcome and Ceven picks the right Tisane calls, fills the parameters, and checks the result.
    • Structured, agent friendly tool schemas so each call runs reliably instead of by guesswork.
    • Rich coverage for reading, writing, and querying your Tisane data, across all 13 of its actions.

  2. Managed auth

    • Built in OAuth with automatic token refresh and rotation.
    • One place to manage, scope, and revoke Tisane access.
    • Per user and per environment credentials instead of shared keys.

  3. Agent optimized design

    • Actions are tuned from real success and error rates so reliability climbs over time.
    • Full execution logs so you always know what ran in Tisane, when, and on whose behalf.
    • The agent pauses and asks when Tisane is unclear instead of plowing ahead.

  4. Enterprise grade security

    • Fine grained access so you control which agents and people can reach Tisane.
    • Least privilege by default, read scopes first and only the writes a workflow needs.
    • A full audit trail of every Tisane action to support review and sign off.

Supported tools

Every action Ceven's agents can run on Tisane, and when to use it.

Analyze Text
Use this to detect sentiment, entities, and topics within a block of text. This is the primary tool for NLU insights.
Calculate Semantic Similarity
Compare two text fragments to get a numeric similarity score between 0 and 1. Use this to find duplicate queries.
Detect Language
Identify the language code of a provided text string. Use this before routing to a specific translation workflow.
Extract Text
Strip HTML, CSS, or JSON markup to get pure decoded text. Use this to clean web scraped data before analysis.
Get Supported Languages
Pull the full list of languages supported by the API. Use this to validate if a detected language can be processed.
Transform Text
Translate text between languages or paraphrase a sentence. Use this to normalize global feedback into one language.
Analyze Sentiment
Extract the emotional tone of a message. Use this to trigger alerts for angry customers.
Extract Topics
Identify the main subjects discussed in a text. Use this to categorize support tickets automatically.
Detect Entities
Pull names, dates, and locations from a text block. Use this to populate CRM fields from a message.
Filter Problematic Content
Scan text for policy violations or toxic language. Use this to flag content for moderation.
Paraphrase Content
Rewrite a text fragment while keeping the original meaning. Use this to simplify technical jargon for users.
Validate UTF8 Text
Check if the input text is properly encoded. Use this before calling the extraction tool to avoid errors.

12 actions · scroll to see them all

Frequently asked questions

Tisane treats each request as a discrete unit of analysis. If a document contains multiple languages, the Detect Language tool will typically return the dominant language based on the character distribution and linguistic markers. For documents with heavy code switching, we recommend splitting the text into smaller segments or paragraphs before sending them to the API. This ensures that the sentiment analysis and topic extraction are applied using the correct linguistic model for each section, which prevents the accuracy degradation that happens when a single model tries to process a hybrid language string in one go.
Tisane uses advanced natural language processing to identify sentiment, but sarcasm remains a known challenge for all NLU tools. While it can detect negative sentiment markers and contradictory word choices, it may not always label a sarcastic comment as negative if the words used are superficially positive. We suggest combining the sentiment score with the semantic similarity tool to compare the comment against known sarcastic patterns in your specific industry. By building a library of common sarcastic phrases, the Ceven agent can apply a secondary logic layer to catch these nuances more effectively.
Tisane imposes specific character limits per API call depending on your current subscription tier. If you attempt to send a massive document, the API will return an error indicating the payload is too large. To handle this, Ceven automatically implements a chunking strategy where long texts are broken into smaller, overlapping segments. Each segment is analyzed individually, and the agent then aggregates the results to provide a global sentiment and topic summary. This prevents data loss and ensures you stay within the rate limits and payload constraints of your specific Tisane account tier.
The semantic similarity tool returns a floating point number between 0 and 1, where 1 indicates that two pieces of text are semantically identical and 0 indicates no relation. It is not a simple keyword match but a vector based comparison of the meaning. In your Ceven workflows, you can set a threshold, such as 0.85, to trigger a duplicate ticket alert. Anything above that number is treated as a repeat query. This allows the agent to group tickets that use different words to describe the same problem, which is far more powerful than basic search.
The Extract Text tool is designed specifically for markup content like HTML or JSON and does not process binary files such as PDFs or images. If you send a binary stream, the tool will fail or return garbled text. For a successful workflow, you must first use a separate document parsing tool to convert the file into a UTF 8 text string. Once the content is in a raw text format, the Tisane tool can then efficiently strip away the remaining tags and noise to leave only the clean, human readable text for the analysis agents to process.
Tisane processes data as a service and generally does not use your API calls to train their global models unless you have explicitly opted into a data sharing program. However, data is processed in transit and held in short term memory to generate the response. For users with strict data residency requirements, it is important to check which regional endpoint your API keys are tied to. Ceven passes the data directly to the Tisane endpoint without storing a permanent copy of the analyzed text in our own logs, ensuring that your customer data remains private and ephemeral.
If you submit text in a language that falls outside the supported 27, the Detect Language tool will return an unknown code or the closest match with a very low confidence score. If you then try to use the Transform Text tool for translation, the API may return an error or a poor quality translation. To prevent this, we recommend a guardrail workflow: first call Get Supported Languages, then Detect Language, and only proceed to analysis if the detected code exists in the supported list. This prevents wasted API credits on unsupported content.
Tisane provides general purpose topic extraction based on its pre trained models, meaning you cannot upload a custom taxonomy to change how it identifies a topic. It finds the most statistically significant themes in the text automatically. To get specific categories, you can use the semantic similarity tool in a Ceven workflow. You can define a list of your own categories and their descriptions, then have the agent compare the Tisane extracted topics against your custom list to map them to your internal business categories.

Alternatives to Tisane

Other tools that solve a similar problem. Ceven supports these too, so you can switch or run more than one at once.

Google Cloud Natural Language logoGoogle Cloud Natural LanguageAmazon Comprehend logoAmazon ComprehendAzure Cognitive Service for Language logoAzure Cognitive Service for Language

Try Ceven on your stack

Plug Ceven on top of the tools you already run. Connect Tisane and the rest of your stack, describe the outcome, and its agents handle the work end to end, days of it in minutes.

Get started for free