Difference between revisions of "Glossary"
m |
m |
||
| Line 13: | Line 13: | ||
=== machine learning === | === machine learning === | ||
A class of algorithm that analyse patterns and relationships in data to make determinations and predictions, using the outcomes of these operations to ''learn'' by iterations and improve future accuracy. Machine learning procedures may be | A class of [[#algorithm|algorithm]] that analyse patterns and relationships in data to make determinations and predictions, using the outcomes of these operations to ''learn'' by iterations and improve future accuracy. Machine learning procedures may be [[#supervised|supervised]], requiring human intervention to provide pre-defined examples with which to ''train'' the algorithm, or [[#unsupervised|unsupervised]], where no human pre-processing of the data is required. | ||
== S == | |||
=== supervised === | |||
In the context of machine learning and statistical analysis, an [[#algorithm|algorithm]] that requires human intervention to provide pre-defined examples with which to [[#training|train]] is said to be ''supervised''. In attribution study, a typical supervised procedure may [[#training|train]] an [[#algorithm|algorithm]] on a corpus comprising texts of known authorship to identify [[#feature|features]] with which to test others. | |||
== T == | == T == | ||
| Line 19: | Line 25: | ||
=== text encoding === | === text encoding === | ||
A set of explicit instructions for the computational representation of text. Whereas a human reader of the text of Othello, for example, is familiar with the conventions distinguishing the word | A set of explicit instructions for the computational representation of text. Whereas a human reader of the text of ''Othello'', for example, is familiar with the conventions distinguishing the word "Othello" as it functions as a title, a running header, a speech prefix, or as a reference to the character in stage directions and dialogue, a computer generally requires these distinctions to be made explicit. Textual encoding or "markup" involves annotating or "tagging" to define units of text (from individual letters to entire documents) into categories. Linguistic features (e.g. grammatical part of speech, syntactic function, and so on) are commonly tagged to support natural language processing, and structural features are used to distinguish elements of text and paratext (e.g. title, dialogue, speech prefix, stage direction, prologue). Other kinds of analysis may require additional categories, such as tagging gender or social status. | ||
=== token === | === token === | ||
A concrete instance of a | A concrete instance of a [[#type|type]]. "I came, I saw, I conquered", for example, contains six word-tokens (excluding punctuation): "came", "saw", "conquered", and three instances of "I". | ||
=== type === | === type === | ||
A unique form; cf. | A unique form; cf. [[#token|token]]. "I came, I saw, I conquered", for example, contains four word-types (excluding punctuation): "I", "came", "saw", and "conquered". | ||
== U == | |||
=== unsupervised === | |||
In the context of machine learning and statistical analysis, an [[#algorithm|algorithm]] that requires no human intervention or pre-processing of the data is said to be ''unsupervised''. This way, the [[#algorithm|algorithm]] processes the data without bias. Principal Components Analysis is an example of a widely used unsupervised method. | |||
== W == | == W == | ||
Revision as of 17:16, 3 July 2021
About the Glossary
Placeholder text about how attribution scholarship often uses highly technical language.
Entries
A
algorithm
A set of logical instructions or mathematical rules used to perform calculations and problem-solving operations, typically by a computer.
M
machine learning
A class of algorithm that analyse patterns and relationships in data to make determinations and predictions, using the outcomes of these operations to learn by iterations and improve future accuracy. Machine learning procedures may be supervised, requiring human intervention to provide pre-defined examples with which to train the algorithm, or unsupervised, where no human pre-processing of the data is required.
S
supervised
In the context of machine learning and statistical analysis, an algorithm that requires human intervention to provide pre-defined examples with which to train is said to be supervised. In attribution study, a typical supervised procedure may train an algorithm on a corpus comprising texts of known authorship to identify features with which to test others.
T
text encoding
A set of explicit instructions for the computational representation of text. Whereas a human reader of the text of Othello, for example, is familiar with the conventions distinguishing the word "Othello" as it functions as a title, a running header, a speech prefix, or as a reference to the character in stage directions and dialogue, a computer generally requires these distinctions to be made explicit. Textual encoding or "markup" involves annotating or "tagging" to define units of text (from individual letters to entire documents) into categories. Linguistic features (e.g. grammatical part of speech, syntactic function, and so on) are commonly tagged to support natural language processing, and structural features are used to distinguish elements of text and paratext (e.g. title, dialogue, speech prefix, stage direction, prologue). Other kinds of analysis may require additional categories, such as tagging gender or social status.
token
A concrete instance of a type. "I came, I saw, I conquered", for example, contains six word-tokens (excluding punctuation): "came", "saw", "conquered", and three instances of "I".
type
A unique form; cf. token. "I came, I saw, I conquered", for example, contains four word-types (excluding punctuation): "I", "came", "saw", and "conquered".
U
unsupervised
In the context of machine learning and statistical analysis, an algorithm that requires no human intervention or pre-processing of the data is said to be unsupervised. This way, the algorithm processes the data without bias. Principal Components Analysis is an example of a widely used unsupervised method.
W
word-token
See token.
word-type
See type.