Generative AI systems for interpreting customer service communications
In an example method, a system accesses first data including text transcripts of a plurality of voice calls, and generates, based on the first data, labeled representations of the voice calls using the one or more computerized LLMs. The system generates the labeled representations by determining a plurality of contextual categories associated with the voice calls, segmenting the text transcript into a plurality of transcript segments, and associating each of the transcript segments with a respective one of the contextual categories. Further, the system generates second data representing the labeled representations of the voice calls and stores the second data using the one or more hardware storage devices.
1 . A method for segmenting and labeling text transcriptions using one or more computerized large language models (LLMs), the method comprising:
accessing, by a computer system from one or more hardware storage devices, first data comprising text transcripts of a plurality of voice calls;
generating, using the one or more computerized LLMs by the computer system based on the first data, labeled representations of the voice calls, wherein the one or more computerized LLMs comprise a generative transformer model having at least one of an encoder or a decoder, wherein generating the labeled representations of the voice calls comprises:
determining, using the one or more computerized LLMs, a plurality of contextual categories associated with the voice calls,
segmenting, using the one or more computerized LLMs, the text transcript into a plurality of transcript segments,
labeling, using the one or more computerized LLMs, each of the transcript segments in accordance with a respective one of the contextual categories, and
based on the labeled transcript segments, generating, using the one or more computerized LLMs, the labeled representations of the voice calls, and
storing, by the computer system, the labeled representations of the voice calls using the one or more hardware storage devices.
2 . The method of claim 1 , further comprising generating second data representing the labeled representations of the voice calls, wherein the second data comprises one or more data structures representing an association between each of the transcript segments and the respective one of the contextual categories.
3 . The method of claim 2 , further comprising:
generating, by the computer system, a graphical user interface, wherein the graphical user interface includes at least a portion of the second data; and
causing, by the computer system, the graphical user interface to be presented to a user.
4 . The method of claim 3 , wherein the graphical user interface comprises:
a first graphical element comprising at least a portion of the text transcripts; and
one or more graphical labels overlaid on the first graphical element, wherein each of the graphical labels represents a respective one of the contextual categories.
5 . The method of claim 1 , further comprising:
determining, for each of the contextual categories, an aggregate call time of the voice calls associated with that contextual category.
6 . The method of claim 5 , further comprising:
generating, by the computer system, a graphical user interface representing the aggregate call times of each of the contextual categories, and
causing, by the computer system, the graphical user interface to be presented to a user.
7 . The method of claim 1 , further comprising:
receiving one or more audio recordings of the voice calls, and
generating, using a computerized speech recognition system, at least some of the text transcriptions based on the one or more audio recordings.
8 . The method of claim 1 , wherein the contextual categories comprise at least one of:
a greeting category,
a caller identification category, or
a conversation close category.
9 . The method of claim 1 , wherein the contextual categories comprise one or more categories associated with providing customer service by a first user to a second user.
10 . The method of claim 1 , wherein determining the contextual categories comprises:
causing the one or more computerized LLMs to generate summaries of each of the text transcripts, and
causing the one or more computerized LLMs to determine a plurality of candidate contextual categories based on the summaries.
11 . The method of claim 10 , wherein determining the contextual categories comprises:
clustering the candidate contextual categories into a plurality of clusters, and
selecting the contextual categories based on the clusters.
12 . The method of claim 1 , wherein the contextual categories are determined using a machine learning process.
13 . The method of claim 1 , each of the transcript segments is associated with a respective one of the contextual categories using a machine learning process.
14 . The method of claim 1 , wherein at least one of the encoder or the decoder is configured to apply a computerized attention mechanism over its respective inputs while generating the labeled representations of the voice calls.
15 . The method of claim 1 , wherein the voice calls are conducted between one or more customers of an organization and one or more representations of the organization.
16 . The method of claim 1 , further comprising:
generating graphs representing a customer journey associated with each of the voice calls.
17 . The method of claim 16 , wherein generating the graphs comprises, for each of the voice calls:
determining, a plurality of states of the voice call;
determining, based on the first data, an order of the plurality of states;
generating a data structure representing the plurality of state of the voice call and an order of the plurality of states; and
generating the graphs based on the data structure.
18 . The method of claim 1 , further comprising:
for each of the voice calls:
determining a call type of the voice call; and
generating the labeled representation of the voice call based on the determined call type.
19 . The method of claim 18 , wherein the contextual categories are determined based on the call type.
20 . The method of claim 1 , wherein the one or more hardware storage devices are implemented on the computer system.
21 . The method of claim 1 , wherein the one or more hardware storage devices are implemented on one or more additional computer systems remote from the computer system.
22 . A system, comprising:
at least one processor; and
a memory communicatively coupled to the at least one processor, the memory storing instructions which, when executed by the at least one processor, cause the at least one processor to perform operations comprising:
accessing, from one or more hardware storage devices, first data comprising text transcripts of a plurality of voice calls;
generating, using one or more computerized large language models (LLMs) based on the first data, labeled representations of the voice calls, wherein the one or more computerized LLMs comprise a generative transformer model having at least one of an encoder or a decoder, wherein generating the labeled representations of the voice calls comprises:
determining, using the one or more computerized LLMs, a plurality of contextual categories associated with the voice calls,
segmenting, using the one or more computerized LLMs, the text transcript into a plurality of transcript segments,
labeling, using the one or more computerized LLMs, each of the transcript segments in accordance with a respective one of the contextual categories, and
based on the labeled transcript segments, generating, using the one or more computerized LLMs, the labeled representations of the voice calls, and
storing the labeled representations of the voice calls using one or more hardware storage devices.
23 . One or more non-transitory computer-readable media storing instructions which, when executed by at least one processor, cause the at least one processor to perform operations comprising:
accessing, from one or more hardware storage devices, first data comprising text transcripts of a plurality of voice calls;
generating, using one or more computerized large language models (LLMs) based on the first data, labeled representations of the voice calls, wherein the one or more computerized LLMs comprise a generative transformer model having at least one of an encoder or a decoder, wherein generating the labeled representations of the voice calls comprises:
determining, using the one or more computerized LLMs, a plurality of contextual categories associated with the voice calls,
segmenting, using the one or more computerized LLMs, the text transcript into a plurality of transcript segments,
labeling, using the one or more computerized LLMs, each of the transcript segments in accordance with a respective one of the contextual categories, and
based on the labeled transcript segments, generating, using the one or more computerized LLMs, the labeled representations of the voice calls, and
storing the labeled representations of the voice calls using one or more hardware storage devices.