What it is

The Internet Archive’s TV News Archive is a vast database of US television news broadcasts captured since 2009, with closed-captioning text indexed for full-text search. It is the canonical research-grade source for cross-network analysis of TV news content, including all Trump TV mentions.

URL: https://archive.org/details/tv

Coverage

• All major US cable news networks (CNN, MSNBC, Fox News, Fox Business, CNBC, Bloomberg)
• Major broadcast networks (ABC, CBS, NBC, PBS)
• Some regional and international news (selective)
• Time-coded video segments with closed-captioning text
• 2009-present (continuous since 2009; comprehensive since 2010)

Strengths

• Massive coverage: Approximately 2 million+ TV news shows
• Closed-captioning text: Full-text searchable across all archived broadcasts
• Time-coded video: Each captioned segment links to the exact video timestamp
• API-accessible: Programmatic search and metadata via archive.org APIs
• Free access: Public domain access for research and education
• Cross-network comparison: Can analyze how the same event is covered across networks

Limitations

• Closed-captioning errors: Auto-generated captioning has substantial transcription errors, especially for proper nouns, technical terms, and rapid speech
• Speaker identification limited: Captions don’t always identify who is speaking
• Indirect Trump coverage: TV captures coverage of Trump, not just Trump speaking. Filtering to Trump’s own speech requires careful query construction
• Time-of-day uncertain: Network-broadcast time vs. event time can differ
• No transcripts of advertisements by design
• Subject to take-down requests: Some items have been removed at network request

Ingestion Approach

1. Use the TV News Archive API (https://api.gdeltproject.org/api/v2/tv/tv) — this is the GDELT TV Explorer interface to the IA TV News Archive
2. Query construction for Trump speech:

• Filter to Trump speaking, not coverage of Trump
• Use proximity searches: “Trump” within 5 words of “speaking” or “said”
• Cross-reference with known event dates from APP or Rev
• For interviews: query the network’s interview show (e.g., “Hannity” + Trump)

1. Per-item fields:

• source_id: IA item identifier (TV News Archive ID)
• title: Show title + segment timestamp
• body: Closed-captioning text for the segment (with timestamp markers)
• author: Speaker if identifiable, or unknown_speaker
• published_at: Broadcast date + time
• url: IA TV News Archive segment URL
• doc_type: tv_segment
• extra_metadata: Network, show name, segment timestamp, captioning quality flag (auto_generated_captions), event type if known

1. Mark transcript-quality: All TV-news ingestions should be flagged transcript_quality=auto_generated_captions so analytical pipelines can apply weights or exclude
2. Rate limit: 5 req/sec via the public API; check current limits

Auto-Transcription Quality Notes

Closed-captioning errors typically include:

• Proper-noun substitution (“Trump” → “trumped” or “the trumpet” — common in some shows)
• Compound-word splitting / merging
• Punctuation and sentence-boundary errors
• Transcription dropouts during chaotic scenes (multiple speakers)
• Phonetic errors on uncommon words

For linguistic analysis, TV-news closed-captioning should not be treated as ground-truth verbatim. Use it for:

• Verifying that an event occurred (date+time confirmation)
• Identifying which networks covered which events
• Quantifying coverage volume and topic frequency
• Not for: word-level lexical analysis, sentence-complexity scoring, or precise quotation

TV News Archive Cross-Reference

For verifying any Trump utterance:

1. If APP or Rev has a curated transcript, treat as authoritative
2. If TV News Archive has a captured broadcast of the same event, the video provides ground truth — use it to verify or refine the curated transcript
3. If only TV captioning is available, transcribe the relevant segment manually before quoting

Citation Convention

Internet Archive TV News Archive, “[Show Name],” [Network], [Broadcast Date]. https://archive.org/details/[id]#start=[timestamp].

Maintenance

• Refresh: Weekly via GDELT TV Explorer for new captures
• Backfill: One-time historical pull for 2015-present, prioritizing rallies and interview events
• Quality re-check: Spot-check captioning quality monthly; flag systematic errors per-show for analytical weighting

See Also

• sc-overview.md
• Source Profile — GDELT TV Explorer (the access layer for IA TV News)
• Source Profile — American Presidency Project (UCSB)
• Source Profile — Rev Transcript Library