How Source DNA Works
Source DNA analyzes thousands of articles from each Indian news source to create a unique political fingerprint. Our AI examines word choice, entity coverage, topic selection, and framing patterns to determine where each source falls on the left-center-right political spectrum.
Political Lean Score
Each source receives a score from -10 (strongly left-leaning) to +10 (strongly right-leaning). This is based on actual coverage analysis, not reputation. Scores update daily as new articles are analyzed.
Coverage Fingerprint
Beyond political lean, we visualize each source's DNA across category preferences, sentiment patterns, and topic emphasis. Two sources with the same political lean can have very different coverage fingerprints.
How to Interpret Source DNA Results
DNA Strand Visualization
The DNA strand is a day-by-day heatmap of a source's political coverage. Each vertical bar represents a single day. The top half shows left-lean intensity using blue shading, while the bottom half shows right-lean intensity using red shading. Darker colours indicate a stronger lean on that particular day. A source with alternating dark-blue and dark-red bars has volatile coverage, while a source with uniformly light bars publishes mostly centrist content.
Article Count Breakdown
Alongside the strand, you will see total article counts classified as left-leaning, neutral, or right-leaning. These numbers describe the content of individual articles, not the newspaper's editorial stance. A source may publish articles across the entire spectrum. The breakdown helps you understand whether a source's coverage is predominantly one-sided or evenly distributed.
Categories Analyzed
Political lean is calculated only from Politics, Business, and Social category articles, because these are the categories where political framing is most relevant and measurable. Sports, entertainment, and technology articles are excluded from the lean calculation to avoid noise. This focused approach ensures the lean score reflects genuine editorial tendencies rather than topic mix.
Important Caveats
Source DNA measures the political lean of individual articles, not a newspaper's official position. A source that publishes both left-leaning opinion pieces and right-leaning news reports will show a balanced overall score. The analysis is based on actual content patterns, not reputation, and scores can shift over time as editorial decisions evolve.
Methodology: Political Lean Classification
Our political lean classification uses a multi-factor AI model trained on over 100,000 human-labelled Indian news articles. The model achieves 78% agreement with expert annotators on a three-class (left, centre, right) classification task. Here is how the scoring pipeline works.
Classification Model
Each article is scored on a scale from -10 (strongly left-leaning) to +10 (strongly right-leaning). The model analyses four dimensions: entity framing (how politicians and parties are described), policy stance indicators (support or opposition to specific policy positions), loaded language patterns (words and phrases associated with partisan framing), and source attribution choices (which voices are quoted and how they are introduced). These signals are combined into a single lean score that updates daily as new articles are ingested.
Training and Validation
The classifier was trained on a dataset of Indian news articles with human-validated political lean annotations contributed by journalism researchers. Validation is performed on a held-out test set using stratified sampling to ensure each lean category is adequately represented. We continually monitor classifier drift by comparing automated scores against periodic expert reviews, and retrain when agreement drops below acceptable thresholds.
Research basis: Our political lean detection draws on media bias research (Groseclose & Milyo, 2005; Gentzkow & Shapiro, 2010) and recent NLP approaches (Baly et al., 2018, 2020) for automated political stance detection.
Frequently Asked Questions
How do you determine if a news channel is left or right wing?
We analyze thousands of articles for political framing, entity coverage patterns, topic selection, and sentiment toward political figures. A score of -10 to +10 indicates left to right lean. The analysis is based on actual coverage, not reputation.
Is NDTV left-wing or neutral?
Our AI analysis of NDTV's coverage patterns shows its political lean based on actual reporting. Check the Source DNA tool for current data - political lean can change over time based on coverage patterns and editorial decisions.
What does the Source DNA fingerprint show?
The fingerprint visualizes a source's coverage patterns across dimensions: political lean, category preferences (politics vs sports vs business), sentiment patterns, and coverage consistency. It's like a unique signature of how each source reports news.