Techmining.com Guide
Ask questions of your own data — and get sourced answers
Techmining.com is a Techmining platform, powered by vantagenAI™, delivered as a service. You upload an export of patent, publication, or innovation records, tell it which fields matter, and then ask questions of that data in plain language. The AI reads your actual records and writes back an analysis grounded in them — not generic prose — which you can export to Word or PDF and hand to a stakeholder.
This guide covers the whole workflow, from preparing a file to exporting a finished report, plus credits, connecting AI agents, privacy, and troubleshooting. Keep it open in a separate tab while you work if you like.
1. The workflow at a glance
Every analysis follows the same five steps:
- 1 Upload. Bring an export you already have. Techmining.com reads its structure and proposes the fields worth analyzing.
- 2 Confirm the fields. Tell it which columns hold your text, organizations, dates, and countries, so nothing is guessed.
- 3 Ask. Generate a dataset summary, pick a guided Techmining question, or type your own question in plain language.
- 4 Review the answer. The AI writes an analysis drawn from your actual records. Follow up in the same conversation to go deeper.
- 5 Export. Send any finding to Word or PDF. Your chat history stays with the dataset for next time.
2. Uploading your data
Start with a text-record export from a search platform such as Lens, PatBase, Scopus, Web of Science, or a spreadsheet you maintain yourself. Techmining.com analyzes the words in your records — titles, abstracts, claims, assignee names — so the more descriptive text a record carries, the more it has to work with.
Supported formats
- Spreadsheets — CSV, TSV, Excel (
.xls,.xlsx,.ods) - Structured data — JSON, JSONL/NDJSON, XML (including nested exports)
- Field-tagged bibliographic records — RIS,
.nbib,.ciw, BibTeX (.bib) - Plain text —
.txt, and platform bundles such as.pdtf - Archives —
.zip,.7z,.gzare unpacked automatically
What isn't supported
PDFs and image files aren't accepted — there are no text records inside for the model to read. Tables of purely numeric statistics (counts and charts with no descriptive text) also give the analysis little to work with; export the underlying records instead.
Not sure your file will work? Use Test Your Data (under the Help menu, when available) to check that a file parses before you commit it to a full run.
3. Choosing the fields that matter
After a file is read, Techmining.com shows the columns it found and proposes which to use. You have the final say — confirm or change which columns hold each kind of information so the model isn't guessing at your schema:
- Text — titles, abstracts, claims, descriptions: the substance the analysis reads.
- Organization — applicants, assignees, affiliations: who is behind each record.
- Date — publication or priority year: the axis for trends and emergence.
- Country — jurisdiction or origin: the axis for geographic breakdowns.
Locking these in once means the same fields get the same treatment on every question you ask against that dataset. You can revisit a dataset's field selection later if you got something wrong.
4. Asking questions
There are three ways to put a question to your data, and you can mix them in the same conversation.
Summarize the dataset
The Summarize dataset button produces a dataset-level overview — key themes, trends, and outliers, plus a handful of suggested follow-up questions to pull on. It's the best first move on a new dataset.
Guided Techmining questions
The category buttons in the prompt toolbar open prepared Techmining question sets — so an analysis starts from a method, not a blank box. The categories are:
Your own questions
Type any question into the chat box in plain language. Press Enter to send and Shift+Enter for a new line. Follow-ups stay in the same conversation, so you can drill in — "focus on the last three years", "just the top five assignees" — and the analysis builds on what came before.
5. How the AI answers
Techmining.com doesn't answer from general knowledge — it answers from your records. For a broad question it reads across the dataset in chunks, analyzes each chunk, then synthesizes those findings into a single grounded answer. Because the analysis is traceable back to the underlying data, the data stays inspectable and the report stands up to scrutiny.
Follow-ups are efficient. When you ask a follow-up in the same conversation, the platform reuses the work it already did reading your records rather than re-reading the whole dataset — so follow-ups come back faster and cost fewer credits than the first question.
Larger datasets take longer, and reports run as background jobs — you can keep working while one is produced. A dataset moves through a few processing states before it's ready to analyze; the workbench shows you where it is.
6. Working with your results
Every report has its own actions (export, copy, delete), and the ⋮ menu beside the chat box covers the whole conversation:
- Export report — download a Word (.docx) or PDF document of the whole conversation or a single report, with or without your questions.
- Copy — copy one question and its report as text.
- Delete — remove one question and its report. You can undo this for a few seconds.
- Clear conversation — start the conversation fresh for the current dataset.
- Text size — make the conversation larger or smaller.
Your chat history is saved per dataset. Come back next week, reopen the dataset, and the conversation is exactly where you left it — so an analysis is something you return to, not a one-shot query.
7. Credits & storage
Analysis is metered by credits. Generating a summary or a report draws down your balance; cheaper follow-ups draw down less. Your remaining balance for the cycle is always visible in the top navigation bar.
- A monthly membership grants a fresh credit allotment each cycle; unused monthly credits don't carry over.
- When your balance reaches zero, new analysis is paused until your next monthly cycle begins.
- Uploaded data counts against a per-user storage quota; remove datasets you no longer need to free space.
Manage your membership and see usage under Billing, and see what's included on the Pricing page.
8. Connect your AI (API & MCP)
Everything you can do in the browser is also reachable programmatically. Techmining.com runs an MCP server so AI assistants and agents can inspect your datasets and generate reports on your behalf, and issues API keys for direct integration. Both surfaces draw on the same account, credits, and storage as the web app.
Programmatic access is currently behind a beta gate. When it's enabled for your account, the Connect your AI page (once you're signed in) walks you through connecting a client or creating a key.
9. Account, access & privacy
- Early access. The platform is currently invite-and-domain gated while in early access. If you can't sign up, that's why — reach out and we'll help.
- Minimal personal data by design. Accounts use a username rather than your real name, and there's no phone number on file.
- Retention. Inactive datasets and dormant accounts are cleaned up on a schedule, so nothing lingers indefinitely.
- Your data. The records you upload are used to answer your questions; see the Privacy Policy for the full picture.
10. Troubleshooting & getting help
A file won't upload or parse. Check it's one of the supported formats above (not a PDF or image), and try Test Your Data to see exactly what Techmining.com reads. If it's a format that should work but doesn't, send it to us via Contact Us — you can attach the file and we'll take a look.
An answer looks thin. Confirm the right text field is selected — if the analysis is pointed at a numeric or ID column instead of titles/abstracts, it has little to read. Re-check your field selection.
Analysis is paused. You've likely hit a zero credit balance or your storage quota. Check the balance in the nav bar and the Billing page.