How to Connect TextRazor to Excel (4 Methods Compared)

The Problem With Getting Workbook Data In and Out of TextRazor

You have an Excel workbook full of raw text — press releases in one worksheet, support tickets in another, marketing copy flagged for review in a third. You need each row processed for entities, topic categories, or relevance scores, and the results written back into adjacent columns in a shape your team can actually query.

TextRazor is good at pulling structure out of unstructured text: named entities, IAB content categories, IPTC topic codes, relationship graphs, relevance scores. But feeding a spreadsheet column into a REST API and writing the results back is more work than the NLP itself. The typical flow is to export the worksheet as CSV, write a script that loops over rows and calls the API, parse the JSON response, decide which fields to keep, then re-import the output and reconcile row order.

Below are the four common ways teams handle this. Only the last one scales.

Method 1: Export and Script

The default Excel path. Export the worksheet to CSV, write a Python or Node script that reads each row, calls the TextRazor API, and writes output back. Or paste cells one at a time into the API explorer, read the nested JSON, and manually populate the adjacent columns.

For a one-off verification, this gets the job done. But TextRazor's response for a single text includes dozens of entities with types, relevance scores, Wikipedia links, and confidence values — and you're picking the right fields by hand. Do that for 100 competitor press releases and you've spent a morning building a fragile data pipeline from CSV exports.

The deeper problem is that NLP signal compounds across rows. One entity mention tells you little. A hundred of them, sorted by frequency and document, starts telling you something about competitive positioning. Manual export-and-script gives you nothing you can easily aggregate, because each run produces a one-off output file.

Method 2: Power Automate

Power Automate can call the TextRazor API through an HTTP action. You trigger on a new row in an Excel table, make the API call with the cell text as the payload, parse the response, and write back to adjacent columns.

Before going further — are you comfortable writing expressions to extract nested array values in Power Automate? Have you configured a custom HTTP connector with an API key header? Do you know what to do when the response returns an entity array of variable length? If any of those feel unfamiliar, skip to Method 3 or 4.

For those who have done this before: the flow works. You configure the HTTP action to call the TextRazor endpoint, pass your API key in the header, send the cell text as the body parameter, then use expression syntax to navigate the response and write selected fields back to the right columns.

One row fires at a time, always.

Sending 100 press releases through a flow means 100 runs, 100 API calls, and 100 individual writes. If you need a frequency table of mentioned organizations across the full workbook, or want to know which IPTC topic appears most often, the flow has no way to give you that. It processed each row without knowing the others exist.

You probably just need the top entities in a usable column and you probably have no idea how to configure a Power Automate expression that navigates a nested JSON array. So you send the task to whoever handles automations on your team, and now you're waiting. And when you switch from entity extraction to IAB classification, someone has to go back in and remap the output fields.

Method 3: The Previous Generation — Connector Add-Ons

Until recently, the best option for repeatable Excel ↔ TextRazor workflows was a category of add-ons that let you define column mappings and saved run configs. You pointed the tool at your input column, specified which response fields you wanted, and saved the template.

That was a genuine step up from one-at-a-time manual work. Runs were reproducible, output was consistent, and your team didn't have to rebuild the field mapping each time.

But you were still responsible for the field selections, the mapping, the conditional handling for rows with no entity results, and the config update whenever you changed taxonomies or output formats. The add-on moved the data — the thinking stayed on you. And when the workbook structure changed, the config broke until someone rebuilt it.

This is the previous generation. It worked, but it required a dedicated operator who knew the tool.

The Easy Way: Using SheetXAI in Excel

There is a different way entirely. SheetXAI is an AI agent that lives inside your Excel workbook. It reads the workbook, understands what you are looking at, and through its built-in TextRazor integration it can run entity extraction, topic classification, or multi-extractor analysis on any column and write the results back in the shape you describe. No API mapping, no connector config, no script to maintain. You just ask.

Example 1: Entity extraction across a worksheet column

For each text in column A of the 'Press Releases' sheet, use TextRazor entity extraction and write the top entity names and types into columns B and C.

SheetXAI calls the TextRazor API for each row, parses the entity array, selects the top results by relevance score, and writes them back — one row per source text.

Example 2: IAB classification with scores

Classify each description in column A of the 'Content' sheet using the TextRazor IAB taxonomy v3.0 and put the top two categories and confidence scores into columns B, C, D, and E.

The pattern: describe the output structure you want in plain language. SheetXAI handles the field parsing and conditional logic for rows that return no confident category match.

Try It

Get the 7-day free trial of SheetXAI and open any Excel workbook with a column of text you need analyzed, then ask it to run TextRazor entity extraction or topic classification on it. The TextRazor integration is included in every SheetXAI plan.