Text-Transcripts#312tier 1experimental liveNew

`tone delta industry rank`

cadence: Quarterlydata: mediumlong onlyshort onlylong short

2022

Frankel, R., Jennings, J., Lee, J. (2022). "Disclosure Sentiment: Machine Learning vs. Dictionary Methods." Review of Accounting Studies. Extends Loughran-McDonald 2011 with cross-sectional peer rank.

Read the paper →

What it checks

Rank each company's earnings-call tone change against sector peers — buy the most-improved, short the most-deteriorated.

Mechanism

Cross-sectional rank of own FinBERT tone-delta within sector cohort (±45d) strips industry-wide tone drift. Top-decile -> LONG; bottom-decile -> SHORT 30-90d.

No production champion data for this family yet. Stats appear once the discovery pipeline promotes at least one strategy with this family tag, or once a multi-family blend that includes it earns a champion slot.

Signal rule

transcript_finbert_scores tone Y/Y delta percentile vs sector peers; top-decile -> LONG, bottom-decile -> SHORT 30/60/90d

Data dependencies

daily_prices
Adjusted-close OHLCV for every US-listed ticker; primary price feed.
transcript_finbert_scores
Worker data table — see services/worker schema.
stocks
Worker data table — see services/worker schema.

Expected edge

Paper alpha: ~3-5% L/S over 90d on top-vs-bottom decile
Paper window: T+1 to T+90d

Frankel-Jennings-Lee 2022: peer-relative FinBERT tone delta ~3-5% L/S spread over 90d on top-vs-bottom decile.

Related families

transcript peer relative toneText-NLP

Frankel-Jennings-Lee (2022 RAST) "Disclosure Sentiment: Machine Learning vs Dictionary Methods": FinBERT-tone residualized against peers in the same industry has 2× the post-earnings drift R² of raw tone. The idea: an absolute tone of +0.2 means very different things for a defensive consumer-staples name (peer baseline ~+0.3) vs an early-stage biotech (peer baseline ~−0.1). Subtracting the peer median isolates firm-specific narrative; long when peer-relative-tone z ≥ +1, short when ≤ −1.

mdna tone deltaFilings

Loughran-McDonald 2011 JF demonstrate the *tone* of 10-K filings — positive vs negative finance-specific lexicon — predicts forward returns. Distinct from filing_text_delta's *uncertainty* lexicon. Effect: bottom-decile tone-delta underperforms top-decile by ~2-3% over 4 weeks, persisting ~12 weeks. Today consumes the cached Item 1 (Business) text; switches to Item 7 (MD&A) when the EDGAR pipeline emits it (tracked in docs/alpha-research/proposals/).

transcript tone driftText-NLP

Loughran-McDonald 2011 + Price et al. 2012: changes in management's linguistic tone on earnings calls predict near-term price drift. Per-call LM tone = (positive − negative) / (positive + negative) on the FULL transcript; signal on quarter-over-quarter delta. Long when tone rose by > 0.05 vs prior call (hold 60-90 days = one quarter), short when tone dropped by > 0.05.

Explore `tone delta industry rank` on alphactor.ai

See which tickers this family is currently firing on, with live signals and rankings.

Open universe with this filter →Browse all 338 families

tone delta industry rank