particle_podcast_get_share_of_voice

Split a person’s, company’s or entity’s podcast mentions, or the lines that say a keyword, by the speakers’ occupations or fields of research.

Split a person’s, company’s or entity’s podcast mentions by who said them: the speakers’ occupations (SOC 2018) or fields of research (ANZSRC 2020). It answers “who is talking about NVIDIA: investors, engineers, journalists?” in one call. For each group you get its mention lines, its share, and the distinct speakers and episodes over a window, optionally bucketed by day, week or month.

• speakers="guests" leaves hosts out.
• level="detailed" splits into detailed occupations or ANZSRC groups instead of SOC major groups or ANZSRC divisions.
• podcast_slug or publisher_slug narrows the split to one show or network.

Shares are of the mentions spoken by someone with an expertise read. A person listed under several codes counts in each, so shares can sum past 100%. Ad reads are left out unless include_ads is set: ad copy is not a speaker’s voice.

Each group’s code is a guest_occupation (or guest_field) value. To read what that group said about an entity, pass it to particle_podcast_find_mentions with the same entity. find_mentions takes no keyword: for a keyword, particle_podcast_search_transcripts with the same keyword_search and the code finds candidate passages. Its guest filters keep segments where such a guest speaks, not lines, so check each matching line’s speaker; host lines counted by the default all-speakers split are not reached that way.

Keywords

keyword_search splits the lines that say a word or phrase instead of an entity’s mentions, so “who is talking about prior authorization?” has a per-speaker answer. A line counts only when its own speaker said every word in it, case-insensitive. Wrap words in double quotes to require them adjacent and in order. Words are matched as spoken: a word with two meanings counts both, and authorization does not match authorizations. Lines in intros, transitions and outros count as an entity’s mentions do, and ad reads stay out unless include_ads is set.

A keyword’s window may span at most 90 days. A keyword said in more than 3,000 episodes in the window, or matching more content than the keyword index will count, returns validation_error with ways to narrow it: a shorter window, podcast_slug or publisher_slug, or a quoted phrase. A keyword over 500 characters, or one with no words (""), returns validation_error too. Finding a broad keyword’s lines reads every transcript that says it, so a call over a busy word can take around 20 seconds; a specific phrase answers in a few.

This is not what particle_podcast_get_episode_timeseries counts with keyword_search and guest_occupation. That count is episode-level and never checks who spoke: a matching segment counts when its episode had such a guest, even if the host said the words.

For an entity’s mention lines themselves, use particle_podcast_find_mentions; for a keyword’s passages, particle_podcast_search_transcripts. For a brand’s mention volume over time and the shows carrying it, use particle_podcast_get_mention_trends.

Inputs

Field	Type	Required	Default	Description
person_slug	string	one of	—	Person slug from particle_person_resolve (e.g. "sam-altman"). Exactly one of person_slug, company_slug, entity_slug or keyword_search is required.
company_slug	string	one of	—	Company slug or domain from particle_company_resolve (e.g. "nvidia"). Resolves to the company’s linked entity.
entity_slug	string	one of	—	Knowledge-graph entity slug from particle_entity_resolve, for anything that isn’t a person or company (e.g. "germany").
keyword_search	string	one of	—	Split the lines that say these words instead of an entity’s mentions: every word in one line, case-insensitive; wrap words in double quotes to require them adjacent and in order. See Keywords.
by	string	no	occupation	Group the speakers by occupation (SOC 2018) or field (ANZSRC 2020 fields of research).
level	string	no	major	Group level: major (SOC major groups or ANZSRC divisions) or detailed (detailed occupations or ANZSRC groups).
speakers	string	no	all	Whose lines count: all (hosts included) or guests (guests and panelists only).
podcast_slug	string	no	—	Only mentions inside this podcast (slug from particle_podcast_resolve).
publisher_slug	string	no	—	Only mentions on podcasts of this publisher (network), by slug.
include_ads	boolean	no	false	Count mentions inside ad reads too.
published_after	string	no	30 days before published_before	Only episodes published on or after this date (YYYY-MM-DD). The window may span at most 365 days, or 90 for a keyword.
published_before	string	no	now	Only episodes published on or before this date (YYYY-MM-DD).
interval	string	no	—	Add each group’s mentions per day, week or month.
limit	integer (1–50)	no	10	Groups to return, most mentioned first.

Passing no subject returns missing_parameter, and passing two returns invalid_input: call once per subject. A window longer than 365 days (90 for a keyword) returns invalid_parameter. A subject slug that matches nothing returns unresolved_reference, so an unknown name is never read as “nobody mentions it”. A keyword names no record, so a keyword nobody said returns an empty split.

Output

A markdown document headed ## Share of voice — <Name> (<slug>) by <occupation|field> (<major|detailed>) (lines saying <keyword> in place of the name for a keyword), then two rows:

• **Window:** — start to end (end excluded). The window is [start, end), so a published_before date shows as the day after it.
• **Mentions:** — the total mention lines, and how many of them were spoken by someone with an expertise read, with that share. Group shares are of the attributed lines.

A ### Groups section follows, with numbered lines formatted N. Title (code) — M mentions, share S%, P speakers, E episodes. With interval, each group gets an indented line of bucket-start count pairs joined by ·. A closing paragraph repeats that shares can sum past 100% and names the guest_occupation (or guest_field) parameter for a group’s lines: with particle_podcast_find_mentions for an entity, or, for a keyword, particle_podcast_search_transcripts and the same keyword_search to find candidate passages whose speakers you then check.

When no mention in the window was spoken by someone with an expertise read, a single paragraph says so and suggests widening the window. A person with no linked knowledge-graph entity gets a note instead of a split.

Sample (company_slug="nvidia", speakers="guests", limit=3, truncated):

## Share of voice — NVIDIA (nvidia) by occupation (major)

**Window:** 2026-09-10 to 2026-10-11 (end excluded)
**Mentions:** 4812 lines, 2967 spoken by someone with an expertise read (61.66%); shares are of those

### Groups

1. Business and Financial Operations Occupations (13-0000) — 1194 mentions, share 40.24%, 412 speakers, 803 episodes
2. Management Occupations (11-0000) — 1071 mentions, share 36.10%, 389 speakers, 731 episodes
3. Computer and Mathematical Occupations (15-0000) — 402 mentions, share 13.55%, 141 speakers, 287 episodes

A person listed under several codes counts in each, so shares can sum past 100%. Read a group's lines with particle_podcast_find_mentions and guest_occupation=<code>.

Sample (keyword_search="\"prior authorization\"", limit=3):

## Share of voice — lines saying "prior authorization" by occupation (major)

**Window:** 2026-09-11 to 2026-10-11 (end excluded)
**Mentions:** 613 lines, 554 spoken by someone with an expertise read (90.38%); shares are of those
### Groups

1. Healthcare Practitioners and Technical Occupations (29-0000) — 186 mentions, share 33.60%, 90 speakers, 103 episodes
2. Management Occupations (11-0000) — 101 mentions, share 18.20%, 49 speakers, 50 episodes
3. Legal Occupations (23-0000) — 85 mentions, share 15.30%, 7 speakers, 7 episodes

Example

Agent calls: particle_podcast_get_share_of_voice {
  "company_slug": "nvidia",
  "by": "occupation",
  "level": "detailed",
  "speakers": "guests",
  "interval": "week"
}

Then read what the leading group said:

Agent calls: particle_podcast_find_mentions {
  "company_slug": "nvidia",
  "guest_occupation": "13-2051"
}

Related

• REST equivalent: GET /v1/podcasts/mentions/share-of-voice.
• Read a group’s mention lines with particle_podcast_find_mentions and its guest_ filters.
• Look up an occupation or field with particle_expertise_resolve.