System and method for call centre management

1. A computer-implemented method for classifying a topic of a communication using a topic classifier model, the method comprising:

receiving text data converted from a voice communication between two individuals;

splitting the text data into two text data portions including a respective text data portion for each individual of the two individuals;

modifying the respective text data portion for each individual by removing stop words and duplicating key words in the respective text data portion;

merging the respective modified text data portion from each individual of the two individuals to form a text corpus;

classifying one or more topics of the text corpus using a topic classifier model;

identifying a purpose of the voice communication based on the one or more topics from the topic classifier model; and

storing the one or more topics from the topic classifier model in association with the voice communication.

2. The computer-implemented method of claim 1 , further comprising:

building a document term matrix based on the text corpus and a frequency of terms that occur in the text corpus; and

wherein classifying the one or more topics of the text corpus comprises determining the one or more topics based on the document term matrix using the topic classifier model.

3. The computer-implemented method of claim 1 , wherein the topic classifier model includes a latent Dirichlet allocation (LDA) model.

4. The computer-implemented method of claim 1 , wherein the one or more topics and a number of the one or more topics are predetermined.

5. The computer-implemented method of claim 1 , wherein the stop words are defined by a predetermined stop words list.

6. The computer-implemented method of claim 1 , wherein the modifying the respective text data portion for each individual comprises one or more of:

tokenizing the respective text data portion by:

splitting the respective text data portion into sentences;

splitting the sentences into words; and

making the words lowercase and removing punctuation;

removing words that have fewer than three characters from the respective text data portion;

lemmatizing one or more verb words in the respective text data portion by changing verb words in past and future tenses into a corresponding present tense; and

stemming one or more words in the respective text data portion by reducing the one or more words to root form.

7. The computer-implemented method of claim 1 , further comprising: linking the topic classifier model to other data sources.

8. The computer-implemented method of claim 7 , wherein the other data sources include at least one of: call centre data, customer demographics, and mobile and website statistics.

9. The computer-implemented method of claim 7 , further comprising: applying principal component analysis to the text data and the other data sources to generate principal components; applying a topic prediction model to the principal components to predict topics an individual would call about.

10. The computer-implemented method of claim 8 , further comprising: applying 5-fold cross validation to the principal components to generate a 5-fold validated training set.

11. The computer-implemented method of claim 1 , further comprising: during a training phase, building the text corpus from training data including multiple communications.

12. The computer-implemented method of claim 1 , further comprising: clustering the two individuals based on the purpose of the voice communication.

13. The computer-implemented method of claim 12 , wherein the clustering includes performing k-means unsupervised machine learning.

14. The computer-implemented method of claim 13 , further comprising: defining customer groups based on the clustering.

15. The computer-implemented method of claim 1 , further comprising: applying the topic classifier model to a call prediction model to predict if an individual will call an organization.

16. The computer-implemented method of claim 15 , wherein the call prediction model includes a random forest.

17. The computer-implemented method of claim 15 , wherein the call prediction model includes a long short-term memory (LSTM) model.

18. The computer-implemented method of claim 1 , wherein the two individuals include a client and an agent at a call centre and the text data is voice-to-text data.

19. A computer system comprising:

a processor; and

a memory in communication with the processor, the memory storing instructions that, when executed by the processor, cause the system to:

receive text data converted from a voice communication between two individuals;

split the text data into two text data portions including a respective text data portion for each individual of the two individuals;

modify the respective text data portion for each individual by removing stop words and duplicating key words in the respective text data portion;

merge the respective modified text data portion from each individual of the two individuals to form a text corpus;

classify one or more topics of the text corpus using a topic classifier model;

identify a purpose of the voice communication based on the one or more topics from the topic classifier model; and

store the one or more topics from the topic classifier model in association with the voice communication.

20. A non-transitory computer-readable medium having computer executable instructions stored thereon for execution by one or more computing devices, that when executed perform:

receiving text data converted from a voice communication between two individuals;

splitting the text data into two text data portions including a respective text data portion for each individual of the two individuals;

modifying the respective text data portion for each individual by removing stop words and duplicating key words in the respective text data portion;

merging the respective modified text data portion from each individual of the two individuals to form a text corpus;

classifying one or more topics of the text corpus using a topic classifier model;

identifying a purpose of the voice communication based on the one or more topics from the topic classifier model; and

storing the one or more topics from the topic classifier model in association with the voice communication.