Methods and systems for generating problem description

1. A method performed by a computing system comprising:

identifying an incoming voice call from a user device to an agent device associated with the computing system;

generating a transcription of the incoming voice call using one or more natural language processing techniques;

extracting a problem description from the transcription, wherein the problem description indicates a topic for the incoming voice call;

estimating, by a first machine learning model, a situation vector from the problem description;

identifying, by a second machine learning model, a pre-existing situation vector that closely matches the estimated situation vector, the second machine learning model being trained using standardized situation descriptions that include a minimum amount of characters; and

retrieving a situation description that corresponds to the identified pre-existing situation vector.

2. The method of claim 1 , further comprising:

training the second machine learning model by:

generating a training data set comprising a plurality of standardized situation descriptions extracted from pre-existing case notes; and

learning, by the second machine learning model, to map each standardized situation description to a vector space, wherein semantically similar situation descriptions are closely mapped in the vector space.

3. The method of claim 2 , wherein generating the training data set comprising the plurality of standardized situation descriptions comprises:

standardizing each situation description by identifying a subset of situation descriptions that include the minimum amount of characters.

4. The method of claim 2 , wherein generating the training data set comprising the plurality of standardized situation descriptions comprises:

filtering the standardized situation descriptions to include a subset of situation descriptions that start with a same set of characters.

5. The method of claim 2 , wherein generating the training data set comprising the plurality of standardized situation descriptions comprises:

anonymizing each standardized situation description by replacing identifying information with a special token.

6. The method of claim 1 , wherein identifying, by the second machine learning model, the pre-existing situation vector that closely matches the estimated situation vector comprises:

accessing a database comprising a plurality of pre-existing situation vectors; and

generating a cosine similarity between each pre-existing situation vector and the estimated situation vector.

7. The method of claim 1 , further comprising:

training the first machine learning model by:

generating a training data set comprising a plurality of pre-existing transcriptions from a data store and a plurality of situation vectors, each situation vector corresponding to a respective pre-existing transcription; and

learning, by the first machine learning model, to estimate a situation vector for a respective pre-existing transcription based on the training data set.

8. A non-transitory computer readable medium having one or more sequences of instructions, which, when executed by a processor, causes a computing system to perform operations comprising:

identifying, by the computing system, an incoming voice call from a user device to an agent device associated with the computing system;

generating, by the computing system, a transcription of the incoming voice call using one or more natural language processing techniques;

extracting, by the computing system, a problem description from the transcription, wherein the problem description indicates a topic for the incoming voice call;

estimating, by a first machine learning model of the computing system, a situation vector from the problem description;

identifying, by a second machine learning model of the computing system, a pre-existing situation vector that closely matches the estimated situation vector, the second machine learning model being trained using standardized situation descriptions that include a minimum amount of characters; and

retrieving, by the computing system, a situation description that corresponds to the identified pre-existing situation vector.

9. The non-transitory computer readable medium of claim 8 , further comprising:

training, by the computing system, the second machine learning model by:

generating a training data set comprising a plurality of standardized situation descriptions extracted from pre-existing case notes; and

learning, by the second machine learning model, to map each standardized situation description to a vector space, wherein semantically similar situation descriptions are closely mapped in the vector space.

10. The non-transitory computer readable medium of claim 9 , wherein generating the training data set comprising the plurality of standardized situation descriptions comprises:

standardizing each situation description by identifying a subset of situation descriptions that include the minimum amount of characters.

11. The non-transitory computer readable medium of claim 9 , wherein generating the training data set comprising the plurality of standardized situation descriptions comprises:

filtering the standardized situation descriptions to include a subset of situation descriptions that start with a same set of characters.

12. The non-transitory computer readable medium of claim 9 , wherein generating the training data set comprising the plurality of standardized situation descriptions comprises:

anonymizing each standardized situation description by replacing identifying information with a special token.

13. The non-transitory computer readable medium of claim 8 , wherein identifying, by the second machine learning model of the computing system, the pre-existing situation vector that closely matches the estimated situation vector comprises:

accessing a database comprising a plurality of pre-existing situation vectors; and

generating a cosine similarity between each pre-existing situation vector and the estimated situation vector.

14. The non-transitory computer readable medium of claim 8 , further comprising:

training, by the computing system, the first machine learning model by:

generating a training data set comprising a plurality of pre-existing transcriptions from a data store and a plurality of situation vectors, each situation vector corresponding to a respective pre-existing transcription; and

learning, by the first machine learning model, to estimate a situation vector for a respective pre-existing transcription based on the training data set.

15. A system comprising:

a processor; and

a memory having one or more instructions stored thereon, which, when executed by the processor, causes the system to perform operations comprising:

identifying an incoming voice call from a user device to an agent device associated with the system;

generating a transcription of the incoming voice call using one or more natural language processing techniques;

extracting a problem description from the transcription, wherein the problem description indicates a topic for the incoming voice call;

estimating, by a first machine learning model of the system, a situation vector from the problem description;

identifying, by a second machine learning model of the system, a pre-existing situation vector that closely matches the estimated situation vector, the second machine learning model being trained using standardized situation descriptions that include a minimum amount of characters; and

retrieving a situation description that corresponds to the identified pre-existing situation vector.

16. The system of claim 15 , wherein the operations further comprise:

training the second machine learning model by:

generating a training data set comprising a plurality of standardized situation descriptions extracted from pre-existing case notes; and

learning, by the second machine learning model, to map each standardized situation description to a vector space, wherein semantically similar situation descriptions are closely mapped in the vector space.

17. The system of claim 16 , wherein generating the training data set comprising the plurality of standardized situation descriptions comprises:

standardizing each situation description by identifying a subset of situation descriptions that include the minimum amount of characters.

18. The system of claim 16 , wherein generating the training data set comprising the plurality of standardized situation descriptions comprises:

filtering the standardized situation descriptions to include a subset of situation descriptions that start with a same set of characters.

19. The system of claim 15 , wherein identifying, by the second machine learning model of the system, the pre-existing situation vector that closely matches the estimated situation vector comprises:

accessing a database comprising a plurality of pre-existing situation vectors; and

generating a cosine similarity between each pre-existing situation vector and the estimated situation vector.

20. The system of claim 15 , wherein the operations further comprise:

training the first machine learning model by:

generating a training data set comprising a plurality of pre-existing transcriptions from a data store and a plurality of situation vectors, each situation vector corresponding to a respective pre-existing transcription; and

learning, by the first machine learning model, to estimate a situation vector for a respective pre-existing transcription based on the training data set.