Inspect matched sentences

Use the Snippet Tool to inspect the matched sentences that drive the measures. A snippet is a matched sentence; surrounding same-speaker context can be shown with it.

Where to inspect

Matched sentences appear in two places. The dataset page has a built-in sentence browser with highlights and a toggle between the sentence alone and its speaker context. The Snippet Tool is the dedicated review surface with the full filter set described below; use it for systematic validation across companies, sectors, and time.

What to inspect

Review sentences that contribute to Exposure first. Then check the sentences that also contribute to Risk, Positive Sentiment, or Negative Sentiment.

Risk and sentiment measures are conditional on topic-matched sentences. A sentence must match the topic query before it can contribute to those measures.

When you open the Snippet Tool, confirm that the selected dataset is the one you intend to review. The tool may default to a recent dataset and date range.

Use filters for review

Filter matched sentences by:

• Date range.
• Headquarters country.
• Sector (economic or business) and SIC sector.
• Company name and earnings-call identifier.
• Measure type: exposure, risk, positive, negative, or sentiment.
• Additional keywords, to show only matched sentences that also contain a given term.
• Latest call per company, to deduplicate the view down to each company's most recent call.

These filters narrow the review after a search. The filters available in the query differ. See Query and dataset settings. Specific companies can be focused on only here or in the dataset export, not before the search.

The sentiment measure filters are stricter than the main measure counts: they show only one-sided positive or one-sided negative sentences and exclude mixed sentences. See the filter caveat before comparing visible snippets against exported counts.

Where highlights are shown, use them to separate the pieces of the match:

• Topic keywords are highlighted separately from dictionary words.
• Risk or uncertainty words identify why a sentence contributes to Risk.
• Negative and positive sentiment words identify why a sentence contributes to the sentiment counts.

Record what you learn

While reviewing matched sentences, record:

• False positives to exclude.
• Missing synonyms or phrase variants.
• Ambiguous terms.
• Useful plural forms.
• Examples that support the final construct definition.

Then refine the query and rerun the search if needed. Keep the notes with the reproducibility record.

What snippets do not prove

Matched sentences can show why a measure value was counted. They do not prove that the topic is economically important, causal, predictive, or complete for all business risk.