# Create language evaluation for Speech Recognition ### Why do you need a Bulk and Evaluation Workflow? When testing ASR models, your main goal is to scope out which model is best suited for your project needs. There are several components to test: * understanding model proficiency in global languages and local accents * assessing the accuracy of translations * checking for low Word Error Rate * latency * price ### Features of Bulk Runner and Evaluation Workflow 1. Run several models in one click 2. Choose any of the API Response Outputs to populate your test 3. Built-in evaluation tool for quick analysis 4. Use csv or Google Sheets as input 5. Get output in CSV for further data analysis #### Also see:


🏎️ Global Language Understanding for AIs	Global Language Understanding for AIs	gooey.ai - cute robots racing vintage painted magazine advertisement muted colorful illustration.png
🗣️ Check out our Hindi ASR Evaluation	https://gooey.ai/bulk/compare-hindi-speech-recognition-hkgs8120p11t/	gooey.ai - cute vintage poster style illustration of a small group of indian kids learning hindi (1).png

## Getting Started ### Step 0 - Prepare your data 1. Collect all your audio voice samples into a Google Drive folder (make sure they are in .wav or .mp3 format) 2. In a new Google spreadsheet or CSV file copy the links of all your audio files 3. Add the human-created transcription for each sample 4. Add the English translation for each sample

### Step 1 - Select the ASR Models Head to our bulk and eval workflow. [LINK TO HINDI ASR EVALUATION EXAMPLE](https://gooey.ai/bulk/compare-hindi-speech-recognition-hkgs8120p11t/) In the example, we have already pre-filled the various models that can be tested. You can choose the ones you want to run by selecting it in the dropdown.

Select the models you want to evaluate your audio samples on

### Step 2 - Add your CSV/Google Sheets Upload your CSV/Google Sheet from [Step 0](#step-0-prepare-your-data). In this example, we have used a Google Sheet of 10 Audio Samples with transcripts and translations. A preview of your sheet will appear once it is correctly uploaded.

### Step 3 - Select the input column Select the column in the input from the dropdown box. The outputs will appear as various columns. In this example, it will be the "audios" column

### Step 4 - Select the pre-built evaluator In the "Evaluation Workflows" section select the "Speech Recognition Model Evaluator".

### Step 5 - Hit Submit Once you hit submit, the selected ASR model workflows (see [Step 1](#step-1-select-the-asr-models)) will run for each audio file in the sheet (see [Step 2](#step-2-add-your-csv-google-sheets)). An output CSV will be generated on the right-hand side of the page.

The outputs will appear in a table format on the right of the page, the output will appear in new columns after your originally populated columns

After the runs are complete, the selected Evaluator (see [Step 4](#step-4-select-the-pre-built-evaluator)), will compare the ASR model outputs to the human-generated translations. It will assess and rate how accurately each model has translated the audio sample. A bar graph with the performance will appear once the entire evaluation is complete.

After the evaluation is complete, table and a bar graph will show the evalution scores

## FAQs #### Q: How should I prepare the transcription and translation data? A: Arrange the audio sample link in the first column, for each audio link add transcriptions and translations in the respective row.

Screenshot of audio sample links with trasncriptions and english translation

#### Q: What is the ideal length of the recording? A: Any recording less than 40 minutes will work successfully, if you are a copilot creator we would recommend limiting the audio sample to 2-3 minutes. #### Q: What audio format should we use? A: Gooey's ASR workflow will accept .wav and .mp3 audio formats. #### Q: How many files can I test? A: Our partners have tested up to 1000 audio files in one run! #### Q: How can we increase the quality of the test? A: Make sure your transcriptions and translations are as accurate as possible.