presidential-speeches — Dataset Codebook
Generated: 2026-11-15
Columns
president
- Type:
object - Non-null: 1,059 / 1,059 (100.0%)
- Description: U.S. president who delivered the speech.
speech_date
- Type:
datetime64[us] - Non-null: 1,059 / 1,059 (100.0%)
- Description: Date of delivery (YYYY-MM-DD).
year
- Type:
int32 - Non-null: 1,059 / 1,059 (100.0%)
- Description: Year of delivery.
title
- Type:
object - Non-null: 1,059 / 1,059 (100.0%)
- Description: Miller Center speech title.
url
- Type:
object - Non-null: 1,012 / 1,059 (95.6%)
- Description: Source URL at millercenter.org.
source_file
- Type:
object - Non-null: 1,059 / 1,059 (100.0%)
- Description: Original JSON filename in the Miller Center archive.
word_count
- Type:
int64 - Non-null: 1,059 / 1,059 (100.0%)
- Description: Total word tokens (transparent lowercase tokenizer, HTML entities decoded).
self_count
- Type:
int64 - Non-null: 1,059 / 1,059 (100.0%)
- Description: Count of self pronouns: I, me, my, mine, myself (+ contractions I’m/I’ve/I’ll/I’d).
collective_count
- Type:
int64 - Non-null: 1,059 / 1,059 (100.0%)
- Description: Count of collective pronouns: we, us, our, ours, ourselves (+ we’re/we’ve/we’ll/we’d/let’s).
self_per_1k
- Type:
Float64 - Non-null: 1,059 / 1,059 (100.0%)
- Description: self_count per 1,000 words.
collective_per_1k
- Type:
Float64 - Non-null: 1,059 / 1,059 (100.0%)
- Description: collective_count per 1,000 words.
self_share
- Type:
Float64 - Non-null: 1,056 / 1,059 (99.7%)
- Description: self_count / (self_count + collective_count). 0.5 = balanced; null if neither present.
speech_type
- Type:
object - Non-null: 1,059 / 1,059 (100.0%)
- Description: Institutional type from the title (Inaugural Address, State of the Union, Annual Message, etc.).
is_sotu_series
- Type:
bool - Non-null: 1,059 / 1,059 (100.0%)
- Description: True if State of the Union OR (pre-1929) Annual Message — the unified SOTU series.
delivery_mode
- Type:
object - Non-null: 1,059 / 1,059 (100.0%)
- Description: written_era (1801-1912 clerk-read messages) vs spoken_era — a comparability break.
Notes
Source: Miller Center (UVA) presidential speech archive, public domain. CURATED corpus of major speeches (not exhaustive); coverage denser for modern presidents. Measures the speech AS DELIVERED (many were ghostwritten). Pronoun/word metrics from a transparent tokenizer (see src/prepare.py).