Terms and Definitions

The following terms were all defined in Chinchor (2000), unless otherwise specified.

• Information Extraction - IE is the process of extracting user-specified text from a set of documents. The extracted text can range from a list of names to a database of event descriptions.
• Information Retrieval - IR is the process of determining the relevant documents from a collection of documents, based on a query presented by the user.
• Automatic Text Summarization - Text summarization is the process of distilling the most important information from a source to produce an abridged version for a particular user and task.
• Machine Translation - MT is the process of rendering text in the source language into text in the target language.
• Topic Detection and Tracking - TDT is the process of segmenting, detecting, and tracking topical information in an information stream. It is essentially a combination of Information Retrieval (detection and tracking) and Speech Recognition (segmentation) technologies (Pallett et al. 2000).
• Speech Transcription - Automated speech transcription is the process of automatically providing a textual representation from audio input. It is sometimes referred to as "automatic speech recognition", or ASR.

Definitions

• Insertion - this is an error by ASR systems in which an extraneous letter or word is inserted into the produced transcript.
• Deletion - this is an error by ASR systems in which a word or letter is erroneously deleted from the produced transcript.
• Substitution - this is an error by ASR systems in which an incorrect word is substituted in the transcript for the correct word from the audio source.
• F-score - this is a metric used most often in the area of IE, but also in the field of Summarization, which combines the two values for Recall and Precision into a single score. This F-score is also referred to as F-measure.
• Recall - recall is a term which describes how many documents are retrieved for a given query, as a percentage of the total number of documents.
• Precision - precision is a value which signifies how many of the documents which were retrieved, were actually relevant to the query.

This page has been accessed - - times since November 14, 2000