Voice Transcript Content by Topic

[ndayGT]

Hello,

I would like to query voice transcripts using /api/v2/speechandtextanalytics/conversations/ and filter to only include conversations categorized with a specific topic.

by topic I mean the AI / ML ran against the transcript, in addition to sentiment.

is there an API endpoint that returns the topics for a conversation?

Thank you!

[Jerome.Saint-Marc]

Hello,

As far as I know (I haven't practiced transcript search often I must say), you would have 2 different API endpoints to search voice transcripts (or transcripts in general).
There is one API endpoint that you can use to filter per topic, and a second API endpoint that you can use to filter per sentimentScore value/range.
At this time, there is no API endpoint which allows to include both filters using the same API endpoint (I mean using topicId and sentimentScore in the same endpoint filters).

1 - Per Topic ID

If you are trying to retrieve conversations which were categorized with a specific topic, you can use POST /api/v2/analytics/transcripts/aggregates/query. It is similar to what I described in this other post.
You could just add the conversationId in groupBy (to retrieve the conversationId of the conversations with this topic).
Ex:

{
   "interval": "2021-07-07T22:00:00.000Z/2021-07-08T22:00:00.000Z",
   "groupBy": ["mediaType","conversationId"],
   "metrics": ["nTopicCommunications"],
   "flattenMultivaluedDimensions": false,
    "filter": {
        "type": "and",
        "predicates": [
            {
                "dimension": "resultsBy",
                "value": "communication"
            },
            {
                "dimension": "topicId",
                "value": "THE_TOPICID_YOU_ARE_SEARCHING_FOR"
            }
        ]
    }
}

With Python SDK, this would correspond to post_analytics_transcripts_aggregates_query

2 - Per SentimentScore

If you are trying to retrieve conversations with a specific sentimentScore (or range), you could use POST /api/v2/speechandtextanalytics/transcripts/search.
There is an example of search with sentimentScore range on this page.

With Python SDK, this would correspond to post_speechandtextanalytics_transcripts_search

3 - Retrieving SentimentScore and Topics

Once you have the conversationIds (as results of one of the search above):

You will get access to the sentimentScore (and some metrics) using GET /api/v2/speechandtextanalytics/conversations/{conversationId}.
See here for more info.

If you need to access the list of topics for a conversation, you have to use GET /api/v2/speechandtextanalytics/conversations/{conversationId}/communications/{communicationId}/transcripturl and download the transcript itself via the obtained downloadUrl.
See here for more info.
The communicationId corresponds to the id of the call/chat/... of the first participant (customer) that appears in the conversation context (obtained via GET /api/v2/conversations/{conversationId}). See this other post for more details.

Regards,

[ndayGT]

No joke, this is the best reply I've ever received to a question on a forum.

Thank you!

[Jerome.Saint-Marc]

Thank you. Appreciated