Question analysis method, device, knowledge base question answering system and electronic equipment
The present disclosure discloses a question analysis method, a device, a knowledge base question answering system and an electronic equipment. The method includes: analyzing a question to obtain N linearized sequences, N being an integer greater than 1; converting the N linearized sequences into N network topology maps; separately calculating a semantic matching degree of each of the N network topology maps to the question; and selecting a network topology map having a highest semantic matching degree to the question as a query graph of the question from the N network topology maps. According to the technology of the present disclosure, the query graph of the question can be obtained more accurately, and the accuracy of the question to the query graph is improved, thereby improving the accuracy of question analysis.
1. A question analysis method, comprising:
analyzing a question to obtain N linearized sequences, N being an integer greater than 1;
converting the N linearized sequences into N network topology maps;
separately calculating a semantic matching degree of each of the N network topology maps to the question; and
selecting a network topology map having a highest semantic matching degree to the question as a query graph of the question from the N network topology maps;
wherein the separately calculating the semantic matching degree of each of the N network topology maps to the question comprises:
acquiring a semantic representation vector of the question;
acquiring a semantic representation vector of each of the N network topology maps; and
separately calculating the semantic matching degree of each of the N network topology maps to the question according to the semantic representation vector of the question and the semantic representation vector of each of the N network topology maps;
wherein the method further comprises:
in the acquiring the semantic representation vector of the question and the semantic representation vector of the network topology map, exchanging information between the question and the network topology map based on an attention mechanism, to generate the semantic representation vector of the question and the semantic representation vector of the network topology map.
2. The method of claim 1 , wherein the acquiring the semantic representation vector of the question comprises:
acquiring a semantic representation vector corresponding to a word sequence of the question; or
converting the word sequence of the question into a graph structure, and acquiring a semantic representation vector of the graph structure.
3. The method of claim 2 , wherein the graph structure is a fully connected graph, where any word in the word sequence of the question is regarded as a node, and any two nodes are connected.
4. A question analysis device, comprising:
an analysis module, configured to analyze a question to obtain N linearized sequences, N being an integer greater than 1;
a conversion module, configured to convert the N linearized sequences into N network topology maps;
a calculation module, configured to separately calculate a semantic matching degree of each of the N network topology maps to the question; and
a selection module, configured to select a network topology map having a highest semantic matching degree to the question as a query graph of the question from the N network topology maps;
wherein the calculation module includes:
a first acquisition sub-module configured to acquire the semantic representation vector of the question;
a second acquisition sub-module configured to acquire the semantic representation vector of each of the N network topology maps; and
a calculation sub-module configured to separately calculate a semantic matching degree of each of the N network topology maps to the question according to the semantic representation vector of the question and the semantic representation vector of each of the N network topology maps;
the question analysis device further includes: an interaction module, configured to, in the acquiring the semantic representation vector of the question and the semantic representation vector of the network topology map, exchange information between the question and the network topology map based on an attention mechanism, to generate the semantic representation vector of the question and the semantic representation vector of the network topology map.
5. A question analysis device, comprising:
a translation model, configured to acquire a question and analyze the question to obtain N linearized sequences, N being an integer greater than 1;
a sequence-to-graph conversion model, wherein an input end of the sequence-to-graph conversion model is connected to an output end of the translation model, and the sequence-to-graph conversion model acquires the N linearized sequences and converts the N linearized sequences into N network topology maps respectively;
an encoding network, wherein an input end of the encoding network is connected to an output end of the sequence-to-graph conversion model, and the encoding network acquires the question and the N network topology maps; the encoding network performs a first coding on the question, to obtain a semantic representation vector of the question; and the encoding network also performs a second encoding on each of the N network topology maps, to obtain a semantic representation vector of each of the N network topology maps; and
a matching network, wherein an input end of the matching network is connected to an output end of the encoding network, and the matching network acquires the semantic representation vector of the question and the semantic representation vector of each of the N network topology maps, and calculates a semantic matching degree of each of the N network topology maps to the question according to the semantic representation vector of the question and the semantic representation vector of each of the N network topology maps;
wherein the coding network comprises:
a first encoder acquiring and encoding the question, to obtain the semantic representation vector of the question; and
a second encoder, wherein an input end of the second encoder is connected to an output end of the sequence-to-graph conversion model, and the second encoder acquires the N network topology maps and encodes each of the N network topology maps, to obtain the semantic representation vector of each of the N network topology maps;
wherein both an output end of the first encoder and an output end of the second encoder are connected to the input end of the matching network;
wherein the first encoder is a serialization encoder, and the second encoder is a first graph neural network encoder, or, the first encoder is a second graph neural network encoder, and the second encoder is a third graph neural network encoder; or, the first encoder and the second encoder exchange information based on an attention mechanism.
6. A knowledge base question answering system, wherein the knowledge base question answering system comprises the question analysis device of claim 4 .
7. A knowledge base question answering system, wherein the knowledge base question answering system comprises the question analysis device of claim 5 .
8. An electronic device, comprising:
at least one processor; and
a memory communicatively connected with the at least one processor,
wherein the memory stores an instruction executable by the at least one processor, and the instruction are executed by the at least one processor so that the at least one processor is capable of executing the method of claim 1 .
9. A non-transitory computer-readable storage medium storing computer instruction, wherein the computer instruction is configured to allow the computer execute the method of claim 1 .