— iIT-Services

Archive
Tag "Research"

noScribe is an AI-based software designed for transcribing audio, particularly useful for qualitative social research and journalistic interviews. The software is free, open-source (licensed under GPL-3.0), and operates entirely offline, meaning no data is sent to the cloud, ensuring privacy. It can recognize up to 99 languages and distinguish between different speakers, which is particularly helpful for interviews with multiple participants.

The software includes an editor that allows users to review, verify, and correct the transcriptions manually. It utilizes advanced AI models, such as OpenAI’s Whisper, faster-whisper by Guillaume Klein, and pyannote by Hervé Bredin, for the transcription process.

It requires a relatively up-to-date computer to function efficiently, slower systems may result in long transcription times. The software is around 3.7 GB, poor audio quality may lead to transcription errors.

noScribe aims to reduce the difficulty of transcription for researchers and journalists, offering a reliable, private, and easy-to-use tool for processing interviews.

Find a review of noScribe (in German) here: https://sozmethode.hypotheses.org/2315.

Source: https://github.com/kaixxx/noScribe

Read More

In collaboration with the Public Prosecutor’s Office of the Canton of Zurich, the Zurich Cantonal Police, the Zurich City Police and the Winterthur City Police, a team from the Statistical Office of the Canton of Zurich has developed a prototype app that automatically transcribes confidential audio and video files. Also in Swiss German.

The app is open-source and based on the Whisper v3 Large model, which enables transcriptions up to 15 times faster than in real time – without license or usage costs. The app offers a wide range of functions such as automatic speaker recognition, multi-file upload, predefined vocabulary and various export options. Transcripts can be edited directly in the application and linked synchronously with the source file.

Hardware requirements: recommend using a CUDA-compatible graphics card with at least 8GB VRAM, as transcription on a CPU is extremely slow.

Source: https://github.com/machinelearningZH/audio-transcription.

Read More

Automatische Transkription, mit Schweizer Mundart-Erkennung und Branchenfokus: recapp.ch.

Der Dienst töggl ist ein KI-getriebenes Spracherkennungsprogramm mit Fokus auf Schweizer Sprachen, insbesondere Schweizerdeutsch. Satzzeichen werden von selbst gesetzt. Segmentierung nach Sprechern erfolgt ebenfalls automatisch. Nach dem Transkribieren kann der Text manuell überarbeitet und anschliessend exportiert werden: xn--tggl-5qa.ch, töggl.ch.

Review by Digitec “Schwiizertüütsch transkribieren – Töggl im Test”: digitec.ch/de/page/schwiizertueuetsch-transkribieren-toeggl-im-test-22436.

Trint’s AI turns audio & video files to text in 30+ languages. Tell stories faster by transcribing, translating, editing and collaborating in a single workflow: trint.com.

Weitere Reviews: blog.clickomania.ch/2023/02/03/transkriptions-tools-im-vergleich/#4-3, blog.clickomania.ch/2021/12/08/toeggl-ch/.

Weiteres Tool: iit-services.ch/stt4sg.

Read More

Research project on a sentence-level transcription engine from Swiss German audio to Standard German text, by FHNW & ZHAW: https://stt4sg.fhnw.ch.

Read More

MindProbe is a server to host online experiments. This server runs the open-source software JATOS, which is compatible with several free software packages for developing online experiments, including OpenSesame/ OSWeb, jsPsych, and Lab.JS. MindProbe is sponsored by ESCOP, with Journal of Cognition as their official journal. MindProbe is free: mindprobe.eu.

Read More

Automated transcripts and translations in 40+ languages: https://sonix.ai/.

Read More

This is a step-by-step manual to configure a Limesurvey poll to show X questions from a (random) selection of Y question groups.

The proof of concept can be accessed here: https://survey.toolsforresearch.com/index.php/345678. The compliation was borrowed from Tony Partner:
https://www.limesurvey.org/forum/can-i-do-this-with-limesurvey/107889-randomly-select-question-groups#145950

The full manual can be found here: https://toolsforresearch.com/limesurvey-randomly-show-x-out-of-y-question-groups and an excerpt of it is copied below.

1) The slider to choose the number of groups (variable x)

Create a ‘Multiple numerical input’ question with Qcode ‘gcount’. The maximum should match the number of random groups within randomization group (‘y’).

Add a subquestion ‘1’. The variable ‘x’ will be stored in ‘gcount_1’.

2) Random groups

Create the random groups: ‘randGroups1’ in randomization group.

3) Equation in each random group

Every random group should start with a hidden question of the equation type.

The value of the equation is the sum of the RelevanceStatuses for all the other groups in the same Randomization group.

4) Questions within a random question group

The trick is that every (visible) question within each random group must have a relevance equation that checks if the equation of the current random group is less than gcount_1 (the ‘x’ variable). The relevance equation is entered in the general options of each question.

Read More