Narrative generation for situation event graphs
Described systems and techniques determine an event graph of a causal chain of events representing a situation within a network, the event graph including event text characterizing at least one event of the causal chain of events. The event graph may then be processed using a large language model that includes at least one topological context adapter that includes a graph adapter and a text adapter, including processing the event graph with the graph adapter and the event text with the text adapter. The at least one topological context adapter may be trained using existing narratives describing past situations, and/or may be trained using worklog data describing past situations and corresponding actions taken to remedy the past situations. Outputs of the graph adapter and the text adapter may be combined to generate a narrative of the situation that explains the causal chain of events and/or instructions to remedy the situation.
1 . A computer program product, the computer program product being tangibly embodied on a non-transitory computer-readable storage medium and comprising instructions that, when executed by at least one computing device, are configured to cause the at least one computing device to:
determine an event graph of a causal chain of events representing a situation within a network, the event graph including event text characterizing at least one event of the causal chain of events;
process the event graph using a large language model that includes at least one topological context adapter that includes a graph adapter and a text adapter, including processing the event graph with the graph adapter and the event text with the text adapter, wherein the at least one topological context adapter is trained using existing narratives describing past situations; and
combine outputs of the graph adapter and the text adapter to generate, from the large language model, a narrative of the situation that explains the causal chain of events.
2 . The computer program product of claim 1 , wherein the instructions are further configured to cause the at least one computing device to:
convert the event graph to a text representation of the event graph for providing to the graph adapter.
3 . The computer program product of claim 2 , wherein the graph adapter includes:
graph embedding layers configured to convert the text representation of the event graph into graph embeddings; and
a graph attention network configured to process the graph embeddings.
4 . The computer program product of claim 3 , wherein the graph embedding layers include a vector feature embedding layer configured to convert node features of each node of the event graph and of proximate topology nodes of a network topology into a shared feature space.
5 . The computer program product of claim 3 , wherein the graph embedding layers include an absolute role embedding layer configured to convert a node role of each node of the event graph and of proximate topology nodes of a network topology into a shared feature space.
6 . The computer program product of claim 3 , wherein the graph embedding layers include a relative positional embedding layer configured to convert a relative position of each node of the event graph and of proximate topology nodes of a network topology into a shared feature space.
7 . The computer program product of claim 3 , wherein the graph embedding layers include a hop embedding layer configured to convert a hop distance between each pair of nodes of the event graph and of proximate topology nodes of a network topology into a shared feature space.
8 . The computer program product of claim 1 , wherein the text adapter includes a low rank adapter.
9 . The computer program product of claim 1 , wherein the instructions are further configured to cause the at least one computing device to:
train the at least one topological context adapter including freezing weights of the large language model while updating weights of the at least one topological context adapter using the existing narratives.
10 . The computer program product of claim 1 , wherein the instructions are further configured to cause the at least one computing device to:
combine the outputs of the graph adapter and the text adapter within the at least one topological context adapter using a feed forward neural network.
11 . A computer-implemented method, the method comprising:
determining an event graph of a causal chain of events representing a situation within a network, the event graph including event text characterizing at least one event of the causal chain of events;
processing the event graph using a large language model that includes at least one topological context adapter that includes a graph adapter and a text adapter, including processing the event graph with the graph adapter and the event text with the text adapter, wherein the at least one topological context adapter is trained using existing narratives describing past situations; and
combining outputs of the graph adapter and the text adapter to generate, from the large language model, a narrative of the situation that explains the causal chain of events.
12 . The method of claim 11 , further comprising:
converting the event graph to a text representation of the event graph for providing to the graph adapter.
13 . The method of claim 12 , wherein the graph adapter includes:
graph embedding layers configured to convert the text representation of the event graph into graph embeddings; and
a graph attention network configured to process the graph embeddings.
14 . The method of claim 11 , wherein the text adapter includes a low rank adapter.
15 . The method of claim 11 , further comprising:
training the at least one topological context adapter including freezing weights of the large language model while updating weights of the at least one topological context adapter using the existing narratives.
16 . The method of claim 11 , further comprising:
combining the outputs of the graph adapter and the text adapter within the at least one topological context adapter using a feed forward neural network.
17 . A system comprising:
at least one memory including instructions; and
at least one processor that is operably coupled to the at least one memory and that is arranged and configured to execute instructions that, when executed, cause the at least one processor to:
determine an event graph of a causal chain of events representing a situation within a network, the event graph including event text characterizing at least one event of the causal chain of events;
process the event graph using a large language model that includes at least one topological context adapter that includes a graph adapter and a text adapter, including processing the event graph with the graph adapter and the event text with the text adapter, wherein the at least one topological context adapter is trained using existing narratives describing past situations; and
combine outputs of the graph adapter and the text adapter to generate, from the large language model, a narrative of the situation that explains the causal chain of events.
18 . The system of claim 17 , wherein the instructions are further configured to cause the at least one processor to:
convert the event graph to a text representation of the event graph for providing to the graph adapter.
19 . The system of claim 18 , wherein the graph adapter includes:
graph embedding layers configured to convert the text representation of the event graph into graph embeddings; and
a graph attention network configured to process the graph embeddings.
20 . The system of claim 17 , wherein the instructions are further configured to cause the at least one processor to:
combine the outputs of the graph adapter and the text adapter within the at least one topological context adapter using a feed forward neural network.