Detecting extraneous topic information using artificial intelligence models

1. A computer system comprising:

one or more processors;

a memory coupled to the one or more processors and storing sequences of instructions which, when executed by the one or more processors, causes performing:

receiving a plurality of digitally stored call transcripts that have been prepared from digitally recorded voice calls;

using a topic model of an artificial intelligence machine learning system, the topic model modeling words of a call as a function of one or more word distributions for each topic of a plurality of topics, generating an output of the topic model which identifies the plurality of topics represented in the plurality of call transcripts;

for a particular topic of the plurality of topics, computing a first value representing a vocabulary of the particular topic and a second value representing a consistency of the particular topic in two more call transcripts of the plurality of call transcripts which include the particular topic, wherein computing the first value comprises computing an exponent of an entropy value for a particular word distribution corresponding to the particular topic;

based, at least in part, on one or more of the first value or the second value, determining that the particular topic meets a particular criterion;

updating the output of the topic model to remove the particular topic or distinguish the particular topic from other topics of the plurality of topics which do not meet the particular criterion;

sending the updated output of the topic model to a client computing device.

2. The system of claim 1 , wherein determining that the particular topic meets a particular criterion comprises determining that the first value representing the vocabulary of the particular topic is greater than a first threshold value or that the second value representing the consistency of the particular topic is lower than a second threshold value.

3. The system of claim 1 wherein computing the second value comprises computing a burst concentration across a plurality of call-specific word distributions corresponding to the particular topic.

4. The system of claim 1 :

wherein the one or more word distributions for each of the plurality of topics comprises, for each of the plurality of topics, a first distribution for a first party type and a second distribution for a second party type; and

wherein the instructions, when executed by the one or more processors, further cause performance of:

computing, for the particular topic, a deviation value representing deviations between words in the first distribution for the first party type and words in the second distribution for the second party type;

determining that the particular topics meets the particular criterion based, at least in part, on the deviation value as well as one or more of the first value or the second value.

5. The system of claim 4 , wherein determining that the topic meets the particular criterion comprises determining that the first value representing the vocabulary of the particular topic is greater than a first threshold value and the deviation value representing deviations between words in the first distribution for the first party type and words in the second distribution for the second party type is below a second threshold value.

6. The system of claim 1 , further comprising:

identifying, in the plurality of call transcripts, a plurality of words that correspond to the particular topic;

wherein updating the output of the topic model comprises:

removing, from the plurality of call transcripts, the plurality of words that correspond to the particular topic;

generating a second output of the topic model using the plurality of call transcripts without the plurality of words that correspond to the particular topic.

7. The system of claim 1 , wherein the instructions, when executed by the one or more processors, further causes causing display of a graphical user interface depicting the plurality of topics without the particular topic.

8. The system of claim 1 , wherein the particular topic comprises a scripted topic.

9. The system of claim 1 , wherein the particular topic comprises a broad topic.

10. A computer-implemented method comprising:

receiving a plurality of digitally stored call transcripts that have been prepared from digitally recorded voice calls;

using a topic model of an artificial intelligence machine learning system, the topic model modeling words of a call as a function of one or more word distributions for each topic of a plurality of topics, generating an output of the topic model which identifies the plurality of topics represented in the plurality of call transcripts;

for a particular topic of the plurality of topics, computing a first value representing a vocabulary of the particular topic and a second value representing a consistency of the particular topic in two more call transcripts of the plurality of call transcripts which include the particular topic, wherein computing the first value comprises computing an exponent of an entropy value for a particular word distribution corresponding to the particular topic;

based, at least in part, on one or more of the first value or the second value, determining that the particular topic meets a particular criterion;

updating the output of the topic model to remove the particular topic or distinguish the particular topic from other topics of the plurality of topics which do not meet the particular criterion;

sending the updated output of the topic model to a client computing device.

11. The method of claim 10 , wherein determining that the particular topic meets a particular criterion comprises determining that the first value representing the vocabulary of the particular topic is greater than a first threshold value or that the second value representing the consistency of the particular topic is lower than a second threshold value.

12. The method of claim 10 wherein computing the second value comprises computing a burst concentration across a plurality of call-specific word distributions corresponding to the particular topic.

13. The method of claim 10 :

wherein the one or more word distributions for each of the plurality of topics comprises, for each of the plurality of topics, a first distribution for a first party type and a second distribution for a second party type; and

wherein the method further comprises:

computing, for the particular topic, a deviation value representing deviations between words in the first distribution for the first party type and words in the second distribution for the second party type;

determining that the particular topics meets the particular criterion based, at least in part, on the deviation value as well as one or more of the first value or the second value.

14. The method of claim 13 , wherein determining that the topic meets the particular criterion comprises determining that the first value representing the vocabulary of the particular topic is greater than a first threshold value and the deviation value representing deviations between words in the first distribution for the first party type and words in the second distribution for the second party type is below a second threshold value.

15. The method of claim 10 , further comprising:

identifying, in the plurality of call transcripts, a plurality of words that correspond to the particular topic;

wherein updating the output of the topic model comprises:

removing, from the plurality of call transcripts, the plurality of words that correspond to the particular topic;

generating a second output of the topic model using the plurality of call transcripts without the plurality of words that correspond to the particular topic.

16. The method of claim 10 , further comprising causing display of a graphical user interface depicting the plurality of topics without the particular topic.

17. The method of claim 10 , wherein the particular topic comprises a scripted topic.

18. The method of claim 10 , wherein the particular topic comprises a broad topic.