Join and normalize outputs
Connect your export to other data and choose how to account for transcript length.
Choose a join key
Start with the data you want to join to your export.
| Joining to | Use |
|---|---|
| NL Analytics call-level files | earningscallID, the unique identifier for each call. |
| Compustat | gvkey_compustat from Crosswalk to GVKey. Attach the crosswalk to your main export using earningscallID. |
| Other external data | A matching identifier such as tickersymbol, isin, cik, permid, or ric. See the column reference. |
External identifiers can be missing. Check completeness and match quality in your sample, including when using the GVKey crosswalk.
Choose a denominator
For sentence-count measures, you can keep raw counts and control for transcript length, or divide by a sentence count. Choose the denominator according to the text you want to describe.
| Express counts relative to | Divide by |
|---|---|
| The full call | nr_of_sentences: all sentences in the transcript, even when matching was restricted. |
| The selected sections or speakers | nr_of_sentences_filtered: sentences passing section, speaker-affiliation, or speaker-title filters. Present when these restrictions are used. |
Scoring context does not by itself change the sentence denominator or create nr_of_sentences_filtered. Check the scoring mode before interpreting a ratio as a sentence share.
Example: the same count, two shares
In this illustrative call, there are 200 sentences. A speaker filter leaves 50 eligible sentences, of which 10 match the query. Exposure is 10.
- Share of eligible sentences: 10 ÷ 50 = 20%.
- Filtered matches as a share of the full call: 10 ÷ 200 = 5%.
Both calculations use the same 10 matches. The second includes excluded speakers' sentences in the denominator, but does not count their topic matches.
Check the joined sample
Compare the joined data with your intended sample. Check which calls are absent from the export and which were lost in the join; see sample availability. Keep calls with zero matches when they belong in the sample.