Datasets:

qwerttyuiiop
/

FilSwitch

Tasks:

Automatic Speech Recognition

Modalities:

Formats:

Languages:

Size:

Tags:

Libraries:

Dataset card Data Studio Files Files and versions

qwerttyuiiop commited on 5 days ago

Commit

8683445

·

verified ·

1 Parent(s): 74aa96a

Update README.md

Files changed (1) hide show

README.md +1 -2

README.md CHANGED Viewed

@@ -2,7 +2,6 @@
 A **Filipino-English (Taglish) read-speech dataset** created for training and evaluating automatic speech recognition (ASR) systems on code-switched speech. The corpus contains **3,555 manually validated utterances from 152 speakers**, totaling approximately **8.9 hours of audio**.
-|                  |                                    |
 | ---------------- | ---------------------------------- |
 | **Languages**    | Filipino + English code switching  |
 | **Domain**       | Read, Philippine news                               |
@@ -16,7 +15,7 @@ A **Filipino-English (Taglish) read-speech dataset** created for training and ev
 | **Whisper Large v3** |         13.62% |             **9.24%** |             14.71% |
 | MMS                  |         26.41% |            **17.60%** |             27.36% |
 | Qwen3-ASR            |         25.17% |            **20.13%** |             26.45% |
-| Gemma 4              |         16.18% |                     — |                  — |
 Fine-tuning on FilSwitch produced relative WER reductions of **32.16% for Whisper, 33.36% for MMS, and 20.02% for Qwen3-ASR**. In contrast, fine-tuning on Filipino FLEURS did not result in WER reductions on the FilSwitch test set.

 A **Filipino-English (Taglish) read-speech dataset** created for training and evaluating automatic speech recognition (ASR) systems on code-switched speech. The corpus contains **3,555 manually validated utterances from 152 speakers**, totaling approximately **8.9 hours of audio**.
 | ---------------- | ---------------------------------- |
 | **Languages**    | Filipino + English code switching  |
 | **Domain**       | Read, Philippine news                               |
 | **Whisper Large v3** |         13.62% |             **9.24%** |             14.71% |
 | MMS                  |         26.41% |            **17.60%** |             27.36% |
 | Qwen3-ASR            |         25.17% |            **20.13%** |             26.45% |
+| Gemma 4 E4B          |         16.18% |                     — |                  — |
 Fine-tuning on FilSwitch produced relative WER reductions of **32.16% for Whisper, 33.36% for MMS, and 20.02% for Qwen3-ASR**. In contrast, fine-tuning on Filipino FLEURS did not result in WER reductions on the FilSwitch test set.