Data & Tools
pytempo
A Python library for reading Romanian official statistics from the INS TEMPO Online API. Developed at the Center for Interdisciplinary Data Science, Babes-Bolyai University.
- Finds indicators. 1927 of them, searchable by keyword, or filtered by territorial level, domain, periodicity and whether they carry a CAEN classification.
- Reads the metadata. Definitions, methodology, sources and observations exactly as INS wrote them, plus the dimensions with their roles and sizes.
pip install git+https://github.com/CIDS-UBB/pytempo.git
TranscriptionLab
A local desktop application for transcribing and reviewing interviews, with the API key of your choice. The recording is sent to the transcription provider you choose in Settings, the language of the transcription is chosen separately, and the result can be reviewed and exported to DOCX, TXT and JSON.
What it does
- Five transcription providers, switchable in Settings: Gladia, Soniox, Deepgram, OpenAI
gpt-4o-transcribe-diarize, and any OpenAI-compatible endpoint you configure. - Speakers identified across the whole recording with Gladia, Soniox and Deepgram: roughly as many labels as there were real participants.
- Transcribe only a time range of the recording, with timestamps still reported against the full original.
- Review tools: find and replace, reassigning a turn to another speaker, merging two speaker labels, per-turn checked marks and Resume, which reopens the project where you stopped.
- No installer and no administrator rights needed; FFmpeg is included in the download.
Privacy
- The recording goes only to the provider you chose.
- The original file stays on the computer and is never modified.
- API keys stay in the memory of the running process: never saved, logged or exported.
- Temporary files are deleted on reset and on a normal shutdown.