Estudio comparativo de métodos de transcripción para corpus orales: el caso del español
Bibliographic record
Abstract
espanolLos avances tecnologicos han propulsado la metodologia de investigacion en transcripcion. Los programas para corpus linguisticos basados en modelos estadisticos y de aprendizaje profundo han mejorado las fases de alineacion y anotacion. En cambio, cuando se trata de transcribir el material, la carga interpretativa y la propia naturaleza de las conversaciones obstaculizan la automatizacion del proceso. De esta manera, la transcripcion de entrevistas destinadas al estudio de la lengua oral se sigue haciendo con un reproductor y un teclado, y puede convertirse en uno de los aspectos mas largos del procesamiento de datos. Sin embargo, en otros contextos profesionales, el reconocimiento automatico del habla se emplea para transcribir de forma eficaz gracias a la colaboracion humano-computadora. Las tecnicas y estrategias difieren, pero todas tienen en comun que estabilizan las fluctuaciones de las herramientas informaticas y son mas rapidas que otros metodos. En este estudio se ha utilizado una de ellas, el rehablado off-linecon las entrevistas del Corpus oral de la lengua espanola en Montreal. Se ha medido el tiempo empleado, asi como la precision y se ha comparado con el reconocimiento automatico del habla y con la mecanografia. El rehablado off-lineha permitido el uso de un programa automatico de dictado en su estado actual como herramienta para potenciar la transcripcion de entrevistas en menos tiempo y con menos errores. EnglishTechnological advances have propelled the research methodology in transcription. Language corpus tools based on statistical models and deep learning have improved the alignment and annotation phases. However, when it comes to transcribing the material, the conversation’s interpretive load and nature themselves hinder automation of the process. That is why interviews used for studying spoken language are still transcribed with a player and keyboard, which can constitute one of the most time-consuming aspects of data processing. In other professional contexts, automatic speech recognition is used to transcribe effectively through human-computer collaboration. The techniques and strategies may differ, but they all stabilize fluctuations in computing tools and are faster than other methods. In this study, the off-line respeaking method was used to transcribe the interviews of the Spoken Corpus of the Spanish Language in Montreal. Transcription times and accuracy were measured and compared with automatic speech recognition and typing. Off-line respeaking, using automatic speech-to-text software in its current state, proved to be the fastest and most error-free method for transcribing interviews.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.009 | 0.047 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.003 |
| Science and technology studies | 0.002 | 0.001 |
| Scholarly communication | 0.004 | 0.003 |
| Open science | 0.001 | 0.003 |
| Research integrity | 0.002 | 0.001 |
| Insufficient payload (model declined to judge) | 0.007 | 0.003 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".