presidential-readability — Dataset Codebook
Generated: 2026-11-15
Columns
president
- Type:
str - Non-null: 270 / 270 (100.0%)
- Description: President who delivered the speech.
speech_date
- Type:
str - Non-null: 270 / 270 (100.0%)
- Description: Date of the speech (YYYY-MM-DD).
year
- Type:
int64 - Non-null: 270 / 270 (100.0%)
- Description: Calendar year of the speech.
title
- Type:
str - Non-null: 270 / 270 (100.0%)
- Description: Speech title as given by the Miller Center.
speech_type
- Type:
str - Non-null: 270 / 270 (100.0%)
- Description: Institutional type (Inaugural Address, State of the Union, Annual Message, etc.).
is_sotu_series
- Type:
bool - Non-null: 270 / 270 (100.0%)
- Description: True if part of the unified State-of-the-Union series (Annual Message <=1928 + State of the Union 1929+).
delivery_mode
- Type:
str - Non-null: 270 / 270 (100.0%)
- Description: ‘written’ (pre-1913, clerk-read documents) or ‘spoken’ (1913+, delivered oratory) — the readability series break.
url
- Type:
str - Non-null: 265 / 270 (98.1%)
- Description: Miller Center source URL for the speech.
source_file
- Type:
str - Non-null: 270 / 270 (100.0%)
- Description: Filename of the source JSON in the corpus (provenance).
fk_grade
- Type:
Float64 - Non-null: 270 / 270 (100.0%)
- Description: Flesch-Kincaid Grade Level (LEAD metric): approximate U.S. school grade required to read the text. Higher = more complex.
flesch_reading_ease
- Type:
Float64 - Non-null: 270 / 270 (100.0%)
- Description: Flesch Reading Ease (0-100): higher = easier to read.
smog_index
- Type:
Float64 - Non-null: 270 / 270 (100.0%)
- Description: SMOG grade-level readability formula.
gunning_fog
- Type:
Float64 - Non-null: 270 / 270 (100.0%)
- Description: Gunning Fog grade-level readability formula.
coleman_liau
- Type:
Float64 - Non-null: 270 / 270 (100.0%)
- Description: Coleman-Liau grade-level readability formula.
word_count
- Type:
Int64 - Non-null: 270 / 270 (100.0%)
- Description: Word (lexicon) count of the transcript, per textstat.
sentence_count
- Type:
Int64 - Non-null: 270 / 270 (100.0%)
- Description: Sentence count of the transcript, per NLTK punkt (the denominator in words/sentence).
Notes
Source: University of Virginia Miller Center presidential speech corpus (public domain), https://data.millercenter.org/miller_center_speeches.tgz. Readability computed with textstat==0.7.13 (Flesch-Kincaid + companions). CAVEAT: readability formulas were built for written prose; scores on transcribed spoken speech are a consistent index, not a literal grade. Segment trends at the written->spoken (~1913) delivery break. Coverage density varies sharply by era.