fdelucaf commited on
Commit
ccb28c6
·
verified ·
1 Parent(s): 8942643

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +18 -5
README.md CHANGED
@@ -48,6 +48,7 @@ specifically idiomatic expressions at sentence level.
48
  This dataset can be used for the assessment of conversational LLMs on two tasks related with idiomatic language:
49
 
50
  Task 1 (monolingual task): idiom detection in an English sentence.
 
51
  Task 2 (cross-lingual task): sentence translation from English to Spanish.
52
 
53
 
@@ -65,11 +66,11 @@ This dataset is not meant to be used for tasks that differ from the ones specifi
65
 
66
  ### Curation Rationale
67
 
68
- This evaluation dataset was designed and curated by human experts with linguistic knowledge, specifically to assess the ability of
69
  LLMs to process figurative language at sentence level.
70
  With the release of this dataset, we aim to provide a resource for evaluating the capabilities of conversational LLMs
71
  to handle the semantic meanings of multi-word expressions and to distinguish between literal
72
- and idiomatic meanings of a potentially figurative expression (PIE).
73
 
74
 
75
  ### Source Data
@@ -77,7 +78,7 @@ and idiomatic meanings of a potentially figurative expression (PIE).
77
  The sentence dataset is based on an original list of English idioms. This list was curated by the same author as the dataset.
78
  The original English idioms are partly derived from real interactions of the author with native English speakers
79
  and partly extracted from the following websites:
80
- [Amigos Ingleses}](https://www.amigosingleses.com/),
81
  [The idioms](https://www.theidioms.com/),
82
  [EF English idioms](https://www.ef.com/wwen/english-resources/english-idioms/).
83
 
@@ -94,6 +95,18 @@ In order to ensure the quality of the generated sentences, the selected collabor
94
  - Demonstrated high linguistic proficiency attaining at least a C1 level.
95
  - Language professional profile with a linguistic background (English teachers, linguists, translators, and NLP experts).
96
 
 
 
 
 
 
 
 
 
 
 
 
 
97
  ### Annotations [optional]
98
 
99
  <!-- If the dataset contains annotations which are not part of the initial data collection, use this section to describe them. -->
@@ -119,8 +132,8 @@ The dataset does not contain any kind of personal or sensitive information.
119
 
120
  A concerted effort was made to mitigate gender bias within our newly developed resource.
121
  Whenever possible, gender-specific terms were either eliminated or neutralised, a large number of sentences
122
- were reformulated adopting a gender neutral first person plural (``we''/``us''), second person singular or
123
- plural (``you''), or third person plural (``they''). Since the gender neutralisation is not always possible
124
  due to grammatical or syntactical constraints, meticulous attention was devoted to ensuring a representation
125
  of feminine and masculine gender terms as balanced as possible throughout the dataset.
126
 
 
48
  This dataset can be used for the assessment of conversational LLMs on two tasks related with idiomatic language:
49
 
50
  Task 1 (monolingual task): idiom detection in an English sentence.
51
+
52
  Task 2 (cross-lingual task): sentence translation from English to Spanish.
53
 
54
 
 
66
 
67
  ### Curation Rationale
68
 
69
+ This evaluation dataset was designed and curated by human experts with advanced linguistic knowledge, specifically to assess the ability of
70
  LLMs to process figurative language at sentence level.
71
  With the release of this dataset, we aim to provide a resource for evaluating the capabilities of conversational LLMs
72
  to handle the semantic meanings of multi-word expressions and to distinguish between literal
73
+ and idiomatic meanings of a potentially idiomatic expression (PIE).
74
 
75
 
76
  ### Source Data
 
78
  The sentence dataset is based on an original list of English idioms. This list was curated by the same author as the dataset.
79
  The original English idioms are partly derived from real interactions of the author with native English speakers
80
  and partly extracted from the following websites:
81
+ [Amigos Ingleses](https://www.amigosingleses.com/),
82
  [The idioms](https://www.theidioms.com/),
83
  [EF English idioms](https://www.ef.com/wwen/english-resources/english-idioms/).
84
 
 
95
  - Demonstrated high linguistic proficiency attaining at least a C1 level.
96
  - Language professional profile with a linguistic background (English teachers, linguists, translators, and NLP experts).
97
 
98
+ The task definition was kept as simple as possible.
99
+ The collaborators were provided with a spreadsheet extracted from the previously compiled list of idioms
100
+ - containing just the idiom and an empty cell for the sentence, without any additional context -
101
+ and were simply instructed to select a few of them of their choice and to craft a sentences per chosen idiom.
102
+ They were asked to produce sentences representative of natural, spontaneous language use by native English speakers,
103
+ allowing for humorous, personal, or improvised content, provided it resonated authentically with their native speaker experience.
104
+ An example idiom with its corresponding sentence was included as a model in the email body:
105
+
106
+ Idiom: ``to have bigger fish to fry''.
107
+ Sentence: ``I don't have time for your silly stories, I have bigger fish to fry: I have a job interview to prepare for tomorrow!''.
108
+
109
+
110
  ### Annotations [optional]
111
 
112
  <!-- If the dataset contains annotations which are not part of the initial data collection, use this section to describe them. -->
 
132
 
133
  A concerted effort was made to mitigate gender bias within our newly developed resource.
134
  Whenever possible, gender-specific terms were either eliminated or neutralised, a large number of sentences
135
+ were reformulated adopting a gender neutral first person plural ("we"/"us"), second person singular or
136
+ plural ("you"), or third person plural ("they"). Since the gender neutralisation is not always possible
137
  due to grammatical or syntactical constraints, meticulous attention was devoted to ensuring a representation
138
  of feminine and masculine gender terms as balanced as possible throughout the dataset.
139