Spoken QA

<p>Spoken questions automatically generated via Google Translate API for the following&nbsp;datasets (test splits only):</p> <ul> <li>Natural Questions <a href="https://ir-datasets.com/beir.html#beir/nq">https://ir-datasets.com/beir.html#beir/nq</a></li> <li>MS MARCO <a href="https://ir-datasets.com/beir.html#beir/msmarco/dev">https://ir-datasets.com/beir.html#beir/msmarco/dev</a></li> <li>Simple Questions (WikiData)&nbsp;<a href="https://github.com/askplatypus/wikidata-simplequestions">https://github.com/askplatypus/wikidata-simplequestions</a></li> </ul> <p>Script used for&nbsp;<a href="https://github.com/svakulenk0/spoken_qa/blob/main/notebooks/generate_speech.ipynb">speech generation</a></p> <p>The automated&nbsp;transcriptions for all spoken questions were made using&nbsp;the Facebook wav2vec2 large ASR model (wav2vec2-large-960h-lv60-self)</p> <p>More info on the project:&nbsp;<a href="https://github.com/svakulenk0/spoken_qa">github.com/svakulenk0/spoken_qa</a></p> <p>&nbsp;</p> <p>&nbsp;</p>