Document expansion for speech retrieval

By Amit Singhal, Fernando C Pereira. 1999

Methods of document expansion for a speech retrieval document by a recognizer. A database of vectors of automatic transcriptions of documents is accessed and the vectors are truncated by removing all terms that are not recognizable by the recognizer to create truncated vectors. Terms in the vectors are then weighted to associate the truncated vectors with the untruncated vectors. Terms not recognized by the recognizer are then added back to the weighted, truncated vectors. The retrieval effectiveness may then be measured.

Subfields: Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Natural Language Processing

1. Jimmy J. Lin
2. Rodrigo Nogueira
3. James R. Glass
4. Hang Li
5. Jun Xu
6. Hung-yi Lee
7. Wei Yang
8. Kyunghyun Cho

Citations in the corpus

• Pretrained Transformers for Text Ranking: BERT and Beyond (2020, background): “bridge the gap between query and document terms is to enrich document representations. This strategy works well for short texts such as tweets [Efron et al., 2012] and noisy transcriptions of speech [Singhal and Pereira, 1999]. Although not as popular as query expansion techniques, researchers never”
• Document Expansion by Query Prediction (2019, background): “, 2012), which works particularly well for speech (Singhal and Pereira, 1999) and multi-lingual retrieval, where terms are noisy.”
• Query Expansion Techniques for Information Retrieval: a Survey (2017, unknown): “In such a case, expanding the transcription by adding related words greatly improves retrieval effectiveness [244].”
• Spoken Content Retrieval—Beyond Cascading Speech Recognition with Text Retrieval (2015, methodology): “probabilistic latent semantic analysis (PLSA) [260] and latent Dirichlet allocation (LDA) [261], each document vector or document language model can be expanded by assigning non-zero weights in (18) or non-zero probabilities in (21) to those terms not appearing in the document but found semantically”
• Semantic Matching in Search (2014, background): “Document reformulation includes key phrase extraction [144, 175], key word or phrase annotation [196, 72, 201], document expansion [164, 114, 169].”